Information processing systems, programs, and information processing methods

The information processing system enhances response generation by interacting with multiple generative models, improving and selecting the best versions, leading to more coherent and refined outputs.

JP2026136039APending Publication Date: 2026-08-25QUANTUM NEXUS INC
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
JP2025039407
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2025-03-12
Publication Date
2026-08-25

AI Technical Summary

Technical Problem

Existing information processing systems struggle to effectively utilize the unique characteristics of multiple AI generation models to generate coherent and improved responses.

Method used

An information processing system that includes an interaction unit to obtain answers from multiple generative models, improve these answers through a first improvement request, and select the best version based on feedback and selection requests, potentially generating a second improved version by further refinement.

Benefits of technology

Enables the generation of more suitable and refined responses by leveraging the strengths of multiple generative models, resulting in improved output quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2026136039000001_ABST
    Figure 2026136039000001_ABST
Patent Text Reader

Abstract

By utilizing multiple different generative models, we can obtain the appropriate information. [Solution] An information processing system comprising a dialogue unit 101 and a presentation unit 103, wherein the dialogue unit 101 obtains a response from each of a plurality of generation models by inputting a prompt containing a request to generate predetermined information to each of the plurality of generation models, and obtains a first improved version of the response by inputting a prompt containing a first improvement request to one generation model that requests improvement of the response based on a plurality of responses obtained from the plurality of generation models.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to an information processing system, a program, and an information processing method.

Background Art

[0002] Conventionally, a technique for referring to information obtained by a plurality of AI models has been known. Patent Document 1 discloses a technique for calculating the amount of foreign exchange hedging based on the future trend of the exchange rate predicted by a plurality of AI models.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, different generation models have different characteristics, and there has been a desire to obtain results that make use of such characteristics.

[0005] The present invention has been made in consideration of such points, and an object thereof is to provide an information processing system, a program, and an information processing method capable of obtaining appropriate information by using a plurality of different generation models.

Means for Solving the Problems

[0006] The information processing system of the present invention is an information processing system provided with an interaction unit, The interaction unit, for each of a plurality of generation models, obtains an answer from each generation model by inputting a prompt including a generation request for predetermined information, The method is characterized by obtaining a first improved version of the answer by inputting a prompt to one generative model that includes a first improvement request requesting an improvement to the answer based on multiple answers obtained from the multiple generative models.

[0007] The information processing system of the present invention is It also includes a display section, The display unit may also display the first improved version.

[0008] The information processing system of the present invention is The dialogue unit inputs a prompt including the first improvement request to each of the plurality of generative models, thereby obtaining the first improved version from each of the plurality of generative models. The display unit may display one first improved version selected from a plurality of first improved versions.

[0009] The information processing system of the present invention is Further equipped with a selection section, The dialogue unit inputs a prompt to each of the multiple generative models, which includes a first selection request requesting the selection of the best first improved version from among the multiple first improved versions, thereby obtaining first selection information from each generative model indicating the first improved version selected by each generative model. The selection unit selects one first improved version based on a plurality of first selection pieces of information. The display unit may display the first improved version selected by the selection unit.

[0010] The information processing system of the present invention is It further comprises a selection section and a presentation section, The aforementioned dialogue unit, By inputting a prompt including the first improvement request to each of the multiple generative models, the first improvement is obtained from each of the multiple generative models. By inputting prompts to each of the multiple generative models, including a first selection request that requests the selection of the best first improved version from among several first improved versions, and a feedback request that requests the identification of missing information in the first improved version selected in response to the first selection request, the first selection information indicating the first improved version selected by each generative model and the missing information are obtained from each generative model. The selection unit selects one first improved version based on a plurality of first selection pieces of information. The dialogue unit inputs a prompt to the generation model that generated the first improved version selected by the selection unit, which includes an improvement request that requests further improvement of the first improved version selected by the selection unit by supplementing the missing information, thereby obtaining a second improved version which is a further improved version of the first improved version. The presenting unit may also present the second improved version.

[0011] The information processing system of the present invention is It further comprises a selection section and a presentation section, The aforementioned dialogue unit, By inputting a prompt including the first improvement request to each of the multiple generative models, the first improvement is obtained from each of the multiple generative models. By inputting a prompt to each of the plurality of generative models, which includes a first selection request requesting the selection of the best first improved version from the plurality of first improved versions, and a feedback request requesting the identification of missing information in the first improved version selected in response to the first selection request, the first selection information indicating the first improved version selected by each generative model and the missing information are obtained from each model. By inputting a prompt to each of the multiple generation models that includes a second improvement request requesting to further improve the first improved version selected by the selection unit by supplementing the missing information, multiple second improved versions, which are further improved versions of the first improved version, are obtained. By inputting a prompt including a second selection request that requests selection of the best second improved version among the second improved versions into each of a plurality of generation models, second selection information indicating the second improved version selected by each generation model is obtained. The selection unit selects one of the second improved versions based on the plurality of second selection information. The presentation unit may present the second improved version selected by the selection unit.

[0012] The information processing system of the present invention is an information processing system including a dialogue unit and a selection unit, wherein the dialogue unit obtains responses from each of the plurality of generation models by inputting a prompt including a generation request that requests generation of predetermined information to each of the plurality of generation models, obtains selection information indicating the best response from each generation model by inputting a prompt including a selection request that requests selection of the best response among the plurality of responses to each of the plurality of generation models, the selection unit selects one of the best responses based on the plurality of selection information, and the dialogue unit obtains an improved version of the response by inputting a prompt including an improvement request that requests improvement of the response based on the plurality of responses to the generation model that generated the response selected by the selection unit.

[0013] The information processing system of the present invention further includes a presentation unit, and the presentation unit may present the improved version.

[0014] The information processing system of the present invention wherein the dialogue unit obtains the selection information and the deficiency information from each generation model by inputting a prompt including the selection request and a pointing request that requests pointing out of deficiency information in the best response together with the plurality of responses to each of the plurality of generation models. A prompt including the improvement request, which requests the generation model that generated the answer selected by the selection unit to supplement the insufficient information and improve the answer based on the plurality of answers, may be input to obtain the improved version.

[0015] The program of the present invention is a program for causing a computer to function as an interaction unit, where the interaction unit inputs a prompt including a generation request for generating predetermined information to each of a plurality of generation models to obtain answers from each generation model, and inputs a prompt including a first improvement request for requesting improvement of an answer based on the plurality of answers obtained from the plurality of generation models to one generation model to obtain a first improved version of the answer, which is characterized in that.

[0016] The program of the present invention is a program for causing a computer to function as an interaction unit and a selection unit, where the interaction unit inputs a prompt including a generation request for generating predetermined information to each of a plurality of generation models to obtain answers from each of the plurality of generation models, inputs a prompt including a selection request for requesting selection of the best answer among the plurality of answers to each of the plurality of generation models to obtain selection information indicating the best answer from each generation model, the selection unit selects one of the best answers based on the plurality of selection information, and the interaction unit inputs a prompt including an improvement request for requesting improvement of an answer based on the plurality of answers to the generation model that generated the answer selected by the selection unit to obtain an improved version of the answer, which is characterized in that.

[0017] The information processing method of the present invention is an information processing method performed by a computer having a control unit, The control unit inputs a prompt containing a request for the generation of predetermined information to each of the multiple generative models, thereby obtaining a response from each generative model. The control unit inputs a prompt to one generative model that includes a first improvement request requesting an improvement to the answer based on a plurality of answers obtained from the plurality of generative models, thereby obtaining a first improved version of the answer. It is characterized by including.

[0018] The information processing method of the present invention is An information processing method performed by a computer having a control unit, The control unit inputs a prompt to each of the multiple generative models, which includes a generation request that requests the generation of predetermined information, thereby obtaining a response from each of the multiple generative models. The control unit inputs a prompt to each of the multiple generative models, which includes a selection request that requests the selection of the best answer from among the multiple answers, thereby obtaining selection information indicating the best answer from each generative model. The control unit performs the step of selecting one of the best answers based on a plurality of selection pieces of information, The control unit inputs a prompt to the generation model that generated the selected answer, which includes an improvement request that requests an improvement to the answer based on the multiple answers, thereby obtaining an improved version of the answer. It is characterized by including. [Effects of the Invention]

[0019] According to the information processing system, program, and information processing method of the present invention, it is possible to present easily understandable summary information of spoken content. [Brief explanation of the drawing]

[0020] [Figure 1] This is an overall diagram of the information processing system. [Figure 2] This is a sequence diagram showing the processing performed by an information processing device. [Figure 3]This is a schematic diagram illustrating the processing in the generative model. [Figure 4] This is a sequence diagram showing the processing of the information processing device according to the second embodiment. [Figure 5] This is a sequence diagram showing the processing of the information processing device according to the third embodiment. [Figure 6] This is a sequence diagram showing the processing of the information processing device according to the fourth embodiment. [Figure 7] This is a sequence diagram showing the processing of the information processing device according to the fifth embodiment. [Figure 8] The following is an example of an answer. [Figure 9] This figure shows an example of the second improved version.

[0021] Embodiments of the present invention will be described below with reference to the drawings.

[0022] (First Embodiment) Figure 1 is an overall configuration diagram of the information processing system 1 of the first embodiment. The information processing system 1 uses a generative model to obtain information (answers) corresponding to questions from the user. For example, the information processing system 1 generates meeting minutes from the spoken information of participants in a meeting.

[0023] Information processing system 1 comprises an information processing device 10, three generative model server devices 20, and a user device 30. The information processing device 10 is an example of an information processing system. The information processing system 1 only needs to include multiple generation model server devices 20, and the number is not limited to the embodiment. The information processing device 10, the generation model server devices 20, and the user device 30 are connected to each other so as to be able to communicate via a communication network 2 such as the Internet.

[0024] The information processing device 10 is a computer or the like and obtains answers to user questions by interacting with the generation model 210 stored in the generation model server device 20. The information processing device 10 is composed of a computer or the like and includes a control unit 100, a storage unit 110, and a communication unit 120.

[0025] The control unit 100 includes a processor such as a CPU (Central Processing Unit) and controls the operation of the information processing device 10. The storage unit 110 includes, for example, an HDD (Hard Disk Drive), RAM (Random Access Memory), ROM (Read Only Memory), and SSD (Solid State Drive). Furthermore, the storage unit 110 is not limited to one built into the information processing device 10, but may be a storage medium that can be detachably attached to the information processing device 10 (for example, a USB memory). In this embodiment, the storage unit 110 stores various programs executed by the control unit 100 and various data.

[0026] The communication unit 120 includes a communication interface for communicating with external devices wirelessly or via wired connections. The control unit 100 transmits and receives data with the generation model server device 20 and the user device 30 via the communication unit 120.

[0027] The generative model server device 20 is a computer or the like. The generative model server device 20 stores a generative model 210 that takes a question (prompt) as input and outputs an answer. The generative model 210 is a learning model generated by machine learning and generates an answer corresponding to the prompt. The generative model can be a Generative Pre-trained Transformer or a language model based thereon. Examples of generative models 210 include ChatGPT, GPT-3, GPT-4 by OpenAI, and GPT-J by EleutherAI.

[0028] The generative model server device 20 is used for large-scale natural language processing tasks, such as automatically performing machine translation, text generation, document classification, and question answering. The generative model 210 learns from a vast amount of data and then makes predictions for new inputs. The generative model server device 20 hosts the generative model 210, calls the generative model 210 in response to requests, and returns the results. In this embodiment, the generative model server device 20 receives a prompt from the information processing device 10 and generates a response sentence by performing calculations. The generated response sentence is then sent to the information processing device 10.

[0029] Each generative model 210 stored in the multiple generative model server device 20 is a different generative model from each other and generates different answers to the same prompt. Each generative model 210 may be a model generated from different training data or a model with different performance.

[0030] The user device 30 is an information processing device. The user device 30 may also be a portable device such as a smartphone, PC, or tablet. The user device 30 comprises a control unit 310, a storage unit 320, a communication unit 330, an operation unit 340, and a display unit 350. The control unit 310, storage unit 320, and communication unit 330 are the same as the control unit 100, storage unit 110, and communication unit 120 of the information processing device 10, respectively. The communication unit 330 includes a communication interface for communicating with external devices wirelessly or via wired connection. The operation unit 340 accepts user input. Examples of the operation unit 340 include a keyboard and a mouse. The display unit 350 is, for example, a monitor and displays various screens. The operation unit 340 and the display unit 350 may be integrated as a touch panel.

[0031] Next, the configuration of the control unit 100 of the information processing device 10 will be described. The control unit 100 functions as the dialogue unit 101, selection unit 102, and presentation unit 103 by executing the program stored in the storage unit 110. In the following, the processes described as being performed by the dialogue unit 101, selection unit 102, and presentation unit 103 are processes performed by the control unit 100 by executing the program.

[0032] The dialogue unit 101 obtains responses from the generation model server devices 20 by sending prompts to multiple generation model server devices 20 via the communication unit 120. The dialogue unit 101 sends prompts, for example, that are entered in response to user operations in the user device 30. The dialogue unit 101 then obtains responses corresponding to the prompts.

[0033] The selection unit 102 selects one answer from among the answers obtained by the dialogue unit 101 from each of the multiple generative models 210.

[0034] The presentation unit 103 presents the user with information such as answers obtained by the generation model 210. Specifically, the presentation unit 103 transmits the information such as answers to the user device 30 via the communication unit 120 and displays it on the display unit 350. In another example, the user device 30 may be equipped with a speaker, and the presentation unit 103 may output audio data of the information such as answers from the speaker. The processing of each functional unit will be described in detail later.

[0035] Figure 2 is a sequence diagram showing the processing of the information processing device 10. Figure 3 is a schematic diagram illustrating the processing in the generation model. In the following, the generation models 210 of the three generation model server devices 20 will be referred to as generation model A, generation model B, and generation model C.

[0036] When the interactive unit 101 of the information processing device 10 receives a prompt from the user device 30 that includes a generation request to generate an answer to a question, it inputs this prompt into three generation models: generation model A, generation model B, and generation model C (step S100). As shown in Figure 3, each generation model A to C generates an answer to the question in response to the prompt that includes the generation request. The answers generated by generation models A, B, and C are referred to as answer A, answer B, and answer C, respectively. The interactive unit 101 of the information processing device 10 obtains answer A, answer B, and answer C from each generation model A to C (step S102).

[0037] Next, the dialogue unit 101 inputs a prompt to each of the generative models A to C that includes a first improvement request, which requests the model to improve its answer based on the multiple answers obtained from each of the generative models A to C (step S104). As shown in Figure 3, each generative model A to C generates an improved version of the answer based on answers A, B, and C. The improved version obtained here is called the first improved version, and the first improved versions generated by generative models A, B, and C are called the first improved version A, the first improved version B, and the first improved version C, respectively. Next, the dialogue unit 101 obtains the first improved versions A to C from each of the generative models A to C (step S106).

[0038] Next, the dialogue unit 101 inputs a prompt to each generation model A to C, which includes a first selection request that asks the model to select the best first improved version from among the first improved versions A to C (step S108). Each generation model A to C selects the best first improved version from among the first improved versions A to C and generates first selection information indicating the selected first improved version. The first selection information is transmitted to the information processing device 10.

[0039] Then, the dialogue unit 101 obtains first selection information from each of the generative models A to C (step S110). Next, the selection unit 102 selects the best first improved version based on the first selection information obtained from each of the generative models A to C (step S112). The selection unit 102 selects the best first improved version, for example, by majority vote. As shown in Figure 3, suppose that generative models A and B generate first selection information indicating first improved version A, and generative model C generates first selection information indicating first improved version C. In this case, first improved version A is selected as the best first improved version.

[0040] Next, the dialogue unit 101 inputs a prompt to the generation model that has generated the best first improved version, which includes a second improvement request that requests further improvement of the first improved version (step S114). For example, if first improved version A is selected as the best first improved version, a prompt including a second improvement request is input to generation model A. The second improvement request is information that requests further improvement of the best first improved version. The generation model that has received a prompt including a second improvement request obtains the second improved version by improving the best first improved version. The dialogue unit 101 may also input multiple first improved versions to the generation model along with the prompt. This allows the generation model to generate a second improved version that is based on the first improved version.

[0041] Next, the dialogue unit 101 obtains the second improved version from the generation model that produced the best first improved version (step S116). Then, the dialogue unit 101 presents the second improved version to the user (step S118). Specifically, the dialogue unit 101 displays the second improved version on the display unit 350 of the user device 30. This completes the process.

[0042] As described above, in the information processing system 1 of this embodiment, the answers obtained from each of the multiple generative models are used to improve the generative model. This makes it possible to obtain an answer that is more suitable to the question than an answer obtained from a single generative model.

[0043] As a modification of the first embodiment, the dialogue unit 101 may include information in the second improvement request for the best first improved version that requests the supplementation of information that is lacking in the best first improved version.

[0044] In this case, the dialogue unit 101 first includes a feedback request in addition to the first selection request in the prompt to be input to all generative models in step S108. Here, the feedback request is information requesting the identification of missing information (missing information) in the first improved version selected by the generative model in response to the first selection request.

[0045] As a result, in step S110, the dialogue unit 101 obtains not only the first selection information but also the missing information from each generation model. Then, in step S114, the dialogue unit 101 inputs a second improvement request to the generation model that generated the best first improved version, requesting that the first improved version be further improved by supplementing the missing information. As a result, in step S116, the dialogue unit 101 can obtain a second improved version that takes the missing information into consideration.

[0046] Furthermore, the dialogue unit 101 may include information in the second improvement request for the best first improvement that requires each first improvement to incorporate features that make it superior to other first improvements.

[0047] In this case, the dialogue unit 101 first includes a suggestion request in addition to the first selection request in the prompt to be input to all generative models in step S108. Here, the suggestion request is information requesting that the generative model point out the advantages of the first improved version selected in response to the first selection request over other first improved versions.

[0048] As a result, in step S110, the dialogue unit 101 obtains not only first selection information from each generative model, but also superiority information indicating that the selected first improved version is superior to other first improved versions. Then, in step S114, the dialogue unit 101 inputs a second improvement request to the generative model that generated the best first improved version, requesting that it further improve the first improved version by incorporating the superiority information. As a result, in step S116, a second improved version that takes the superiority information into consideration can be obtained.

[0049] In this way, when improving the first improved version using a generative model, a better second improved version can be obtained by referring to the information obtained from comparing it with other first improved versions.

[0050] (Second Embodiment) Next, the differences between the information processing system 1 of the second embodiment and the information processing system 1 of the first embodiment will be mainly described. Figure 4 is a sequence diagram showing the processing of the information processing device 10 of the second embodiment. In Figure 4, the same steps as those of the information processing device 10 of the first embodiment, which were described with reference to Figure 2, are assigned the same step numbers.

[0051] In the processing of the information processing device 10 of the second embodiment, when the dialogue unit 101 selects the first improved version in step S112, it then inputs a prompt including a request for a second improvement to all generation models (generation models A to C) (step S130). As a result, a second improved version, which is an improvement on the best first improved version, is generated in each generation model A to C.

[0052] As described in the modified version of the first embodiment, in step S130, the dialogue unit 101 may input a prompt including a request for feedback along with the second improvement request. This makes it possible to obtain a second improved version that takes into account at least one of the missing information and the superior information.

[0053] Next, the dialogue unit 101 obtains second improved versions from each generation model A to C (step S132). Next, the dialogue unit 101 inputs a prompt to each generation model A to C that includes a second selection request, asking them to select the best second improved version from among the second improved versions A to C (step S134). Each generation model A to C selects the best second improved version from among the second improved versions A to C and generates second selection information indicating the selected second improved version. The second selection information is transmitted to the information processing device 10.

[0054] Then, the dialogue unit 101 obtains second selection information from each generative model A to C (step S136). Next, the selection unit 102 selects the best second improved version based on the second selection information obtained from each generative model A to C (step S138). This process is the same as the process of selecting the best first improved version based on the first selection information (step S112). Next, the presentation unit 103 presents the best second improved version selected by the selection unit 102 to the user (step S140).

[0055] Thus, in the information processing system 1 of the second embodiment, information obtained from each of the multiple generative models is used to improve the generative model, so it is possible to obtain an answer that is more suitable to the question than an answer obtained from a single generative model. Furthermore, in the information processing system 1 of the second embodiment, all generative models each generate a second improved version from the first improved version. Since a generative model that generates the best first improved version cannot always generate the best second improved version, each generative model generates a second improved version, and by selecting the best second improved version from among them, a better second improved version can be obtained.

[0056] (Third embodiment) Next, the differences between the information processing system 1 of the third embodiment and the information processing system 1 of the other embodiments will be mainly described. Figure 5 is a sequence diagram showing the processing of the information processing device 10 of the third embodiment. In Figure 5, the same steps as those of the information processing device 10 of the first embodiment, which was described with reference to Figure 2, are assigned the same step numbers.

[0057] In the processing of the information processing device 10 of the third embodiment, if the first improved version is selected in step S112, the dialogue unit 101 presents the selected first improved version to the user (step S150).

[0058] Thus, in the information processing system 1 of the third embodiment, information obtained from each of the multiple generative models is used to improve the generative model, so it is possible to obtain an answer that is more suitable for the question than an answer obtained from a single generative model.

[0059] As a variation of the third embodiment, the selection unit 102 only needs to select one first improved version from the first improved versions obtained from each of the multiple generation models, and the specific processing for this is not limited to the embodiment. Another example is that the selection unit 102 can accept the selection of one first improved version from the user by displaying each first improved version on the display unit 350 of the user device 30, and the selection unit 102 can select one first improved version according to the user operation. Yet another example is that the selection unit 102 can select the first improved version selected by one generation model as the best first improved version.

[0060] Furthermore, as a modification of the above embodiment, the dialogue unit 101 may input a first improvement request to one generation model, thereby obtaining a first improved version from one generation model, and the presentation unit 103 may present the obtained first improved version.

[0061] (Fourth Embodiment) Next, the information processing system 1 of the fourth embodiment will be described, mainly focusing on the differences from the information processing system 1 of the other embodiments. Figure 6 is a sequence diagram showing the processing of the information processing device 10 of the fourth embodiment. In Figure 6, the same steps as those of the information processing device 10 of the first embodiment, which was described with reference to Figure 2, are assigned the same step numbers.

[0062] The information processing device 10 of the fourth embodiment selects the best answer from among multiple answers obtained from multiple generative models and obtains an improved version from the selected answer.

[0063] As shown in Figure 6, when the dialogue unit 101 obtains responses from each generation model A to C in step S102, it then sends a prompt containing a third selection request to each generation model (step S160). The third selection request is information that requests the selection of the best response from among multiple responses. Each generation model A to C selects the best response from responses A to C and generates third selection information indicating the selected response. The third selection information is sent to the information processing device 10.

[0064] Then, the dialogue unit 101 obtains third selection information from each generative model (step S162). Next, the selection unit 102 selects the best single answer based on the third selection information obtained from each generative model A to C (step S164). The selection unit 102 selects the best answer, for example, by majority vote.

[0065] Next, the dialogue unit 101 inputs a prompt including a third improvement request to the generation model that generated the best answer (step S166). The third improvement request is information requesting that the answer be improved based on multiple answers obtained from each of the multiple generation models. The generation model that generated the best answer generates an improved version of the answer based on the third improvement request. The improved version obtained here is called the third improvement version. The third improvement version is sent to the information processing device 10. The dialogue unit 101 acquires the third improvement version (step S168). Next, the presentation unit 103 presents the third improvement version to the user (step S170).

[0066] Thus, in this information processing system, as in the information processing systems of other embodiments, the responses obtained from each of the multiple generative models are used to improve the generative model, thereby enabling the acquisition of more appropriate responses.

[0067] As a modification of the fourth embodiment, similar to the modification of the first embodiment, in step S160, the dialogue unit 101 may input a prompt to the generation model that includes a request for comment along with a third selection request.

[0068] Here, the feedback request is information that requests the identification of missing information (missing information) in the selected answer. Then, in step S162, the dialogue unit 101 acquires not only the third selected information but also the missing information, and in step S166, the dialogue unit 101 inputs a third improvement request, which requests that the answer be improved by supplementing the missing information, into the generative model that generated the best answer. As a result, in step S168, the dialogue unit 101 can obtain a third improved version that takes the missing information into consideration.

[0069] Furthermore, the request for feedback may also be information requesting that the best answer be pointed out as being superior to other answers. In this case, in step S162, the dialogue unit 101 obtains superiority information along with third selection information, and in step S166, it inputs a third improvement request that asks the user to improve the answer by incorporating the superiority information. As a result, in step S168, the dialogue unit 101 can obtain a third improved version that takes the superiority information into consideration.

[0070] (Fifth embodiment) Next, the differences between the information processing system 1 of the fifth embodiment and the information processing system 1 of the fourth embodiment will be mainly described. Figure 7 is a sequence diagram showing the processing of the information processing device 10 according to the fifth embodiment. In Figure 7, the same numbers are used for the same processing as the information processing device 10 of the fourth embodiment, which was described with reference to Figure 6.

[0071] In the processing of the information processing device 10 of the fifth embodiment, when the best answer is selected in step S164, the dialogue unit 101 inputs a third improvement request to all generation models (step S170). As a result, a third improved version, which is an improved version of the best answer, is generated in each generation model A to C. Note that in step S170, the dialogue unit 101 may input a prompt that includes a suggestion request along with the third improvement request.

[0072] Next, the dialogue unit 101 obtains third-refined versions from each of the generation models A to C (step S172). Next, the dialogue unit 101 inputs a prompt to each of the generation models A to C that includes a fourth selection request, asking them to select the best third-refined version from among the third-refined versions (step S174). Each of the generation models A to C selects the best third-refined version from among the multiple third-refined versions and generates fourth selection information indicating the selected third-refined version. The fourth selection information is transmitted to the information processing device 10.

[0073] Then, the dialogue unit 101 obtains fourth selection information from each generative model A to C (step S176). Next, the selection unit 102 selects the best third improved version based on the fourth selection information obtained from each generative model A to C (step S178). Next, the presentation unit 103 presents the best third improved version selected by the selection unit 102 to the user.

[0074] Thus, in this information processing system, as in the information processing systems of other embodiments, the responses obtained from each of the multiple generative models are used to improve the generative model, thereby enabling the acquisition of more appropriate responses. [Examples]

[0075] Regarding the "Creation of High-Performance Organic Semiconductor Materials and Next-Generation Device Applications by Environmentally Harmonious Coating Processes," when creating proposals for multiple generation models, the information processing system 1 of the first embodiment obtained both a response and a second improved version. Figure 8 shows a portion of the response. Figure 9 shows a portion of the second improved version. In the proposal shown in Figure 9, both the research schedule and budget plan are more detailed than in the proposal shown in Figure 8, confirming that the response content has been improved.

[0076] The embodiments described above are merely examples for carrying out the present invention, and various other embodiments can be adopted. For example, various modifications and changes are possible within the scope of the gist of the present invention as described in the claims, such as applying one modification to another. For example, some of the components of the above embodiments may be omitted, or the order of processing may be changed or omitted.

[0077] In the information processing system, program, and information processing method of this embodiment, which have the configuration described above, the information processing system is an information processing system equipped with an interactive unit 101, the interactive unit 101 obtains a response from each of the multiple generation models by inputting a prompt containing a request to generate predetermined information to each of the multiple generation models, and obtains a first improved version of the response by inputting a prompt containing a first improvement request to one generation model that requests improvement of the response based on the multiple responses obtained from the multiple generation models. In this way, appropriate information can be obtained by using multiple different generation models.

[0078] Furthermore, the information processing system, program, and information processing method of this embodiment may further include a presentation unit 103, which may present the first improved version. This allows the user to confirm the first improved version.

[0079] Furthermore, in the information processing system, program, and information processing method of this embodiment, the dialogue unit 101 may input a prompt including a first improvement request to each of the multiple generation models to obtain the first improved version from each of the multiple generation models, and the presentation unit 103 may present one first improved version selected from the multiple first improved versions. By selecting a first improved version to be presented from among the multiple first improved versions, an appropriate first improved version can be presented.

[0080] Furthermore, in the information processing system, program, and information processing method of this embodiment, a selection unit 102 may be further provided, wherein the dialogue unit 101 inputs a prompt to each of the multiple generation models, which includes a first selection request requesting the selection of the best first improved version from among the multiple first improved versions, thereby obtaining first selection information from each generation model indicating the first improved version selected by each generation model, the selection unit 102 selects one first improved version based on the multiple first selection information, and the presentation unit 103 presents the first improved version selected by the selection unit 102. This makes it possible to select a first improved version using a generation model.

[0081] Furthermore, in the information processing system, program, and information processing method of this embodiment, a selection unit 102 and a presentation unit 103 are further included, and the dialogue unit 101 inputs a prompt including a first improvement request to each of the multiple generation models to obtain a first improved version from each of the multiple generation models, and inputs a prompt including a first selection request requesting the selection of the best first improved version from the multiple first improved versions, and a notification request requesting the notification of missing information in the first improved version selected in response to the first selection request to each of the multiple generation models to obtain first selection information and missing information indicating the first improved version selected by each generation model from each generation model, the selection unit 102 selects one first improved version based on the multiple first selection information, and the dialogue unit 101 inputs a prompt including an improvement request requesting the further improvement of the first improved version selected by the selection unit 102 by supplementing the missing information to the generation model that generated the first improved version selected by the selection unit 102 to obtain a second improved version in which the first improved version has been further improved, and the presentation unit 103 presents the second improved version. By further improving this first revised version, it becomes possible to present more appropriate information.

[0082] Furthermore, in the information processing system, program, and information processing method of this embodiment, a selection unit 102 and a presentation unit 103 are further included, and the dialogue unit 101 inputs a prompt including a first improvement request to each of the multiple generation models to obtain a first improved version from each of the multiple generation models, and inputs a prompt including a first selection request that requests the selection of the best first improved version from the multiple first improved versions, and a notification request that requests the notification of missing information in the first improved version selected in response to the first selection request, thereby providing each generation model with first selection information indicating the first improved version selected by each generation model and missing information. By inputting a prompt to each of the multiple generative models, which includes a second improvement request requesting that the first improved version selected by the selection unit be further improved by obtaining information from Dell and supplementing the missing information, multiple second improved versions, which are further improved versions of the first improved version, can be obtained. By inputting a prompt to each of the multiple generative models, which includes a second selection request requesting the selection of the best second improved version among the second improved versions, second selection information indicating the second improved version selected by each generative model can be obtained. The selection unit 102 may then select one of the second improved versions based on the multiple second selection information, and the presentation unit may present the second improved version selected by the selection unit. By further improving the first improved version in this way, more appropriate information can be presented.

[0083] Furthermore, in the information processing system, program, and information processing method of this embodiment, the information processing system comprises a dialogue unit 101 and a selection unit 102. The dialogue unit 101 obtains answers from each of the multiple generation models by inputting a prompt including a generation request that requests the generation of predetermined information to each of the multiple generation models, and obtains selection information indicating the best answer from each of the multiple generation models by inputting a prompt including a selection request that requests the selection of the best answer from among the multiple answers to each of the multiple generation models, and the selection unit 102 selects one of the best answers based on the multiple selection information, and the dialogue unit 101 obtains an improved version of the answer by inputting a prompt including an improvement request that requests the generation model that generated the answer selected by the selection unit 102 to improve the answer based on the multiple answers. In this way, appropriate information can be obtained by using multiple different generation models.

[0084] Furthermore, the information processing system, program, and information processing method of this embodiment may further include a presentation unit 103, which may present an improved version. This allows the user to review the improved version.

[0085] Furthermore, in the information processing system, program, and information processing method of this embodiment, the dialogue unit 101 may obtain selected information and missing information from each of the multiple generation models by inputting a prompt to each of the multiple generation models, which includes a selection request and a suggestion request that requests the identification of missing information in the best answer, along with multiple answers. The dialogue unit 101 may then input a prompt to the generation model that generated the answer selected by the selection unit 102, which includes an improvement request that requests the model to improve the answer by supplementing the missing information based on the multiple answers. This will allow an appropriate improved version to be obtained. [Explanation of Symbols]

[0086] 1. Information Processing System 10 Information Processing Devices 20 Generative Model Server Device 30 User devices 100 Control Unit 101 Dialogue Section 102 Selection Section 103 Presentation section 110 Storage section 120 Communications Department 210 Generative Models 310 Control Unit 320 Storage section 330 Communications Department 340 Operation section 350 Display section

Claims

1. An information processing system equipped with an interactive section, The aforementioned dialogue unit, By inputting a prompt containing a request to generate specific information for each of the multiple generative models, responses are obtained from each generative model. An information processing system that obtains a first improved version of an answer by inputting a prompt into one generative model that includes a first improvement request requesting an improvement to an answer based on multiple answers obtained from the multiple generative models.

2. It also includes a display section, The information processing system according to claim 1, wherein the display unit displays the first improved version.

3. The dialogue unit inputs a prompt including the first improvement request to each of the plurality of generative models, thereby obtaining the first improved version from each of the plurality of generative models. The information processing system according to claim 2, wherein the display unit displays one first improved version selected from a plurality of first improved versions.

4. Further equipped with a selection section, The dialogue unit inputs a prompt to each of the multiple generative models, which includes a first selection request requesting the selection of the best first improved version from among the multiple first improved versions, thereby obtaining first selection information from each generative model indicating the first improved version selected by each generative model. The selection unit selects one first improved version based on a plurality of first selection pieces of information. The information processing system according to claim 3, wherein the presentation unit presents the first improved version selected by the selection unit.

5. It further comprises a selection section and a presentation section, The aforementioned dialogue unit, By inputting a prompt including the first improvement request to each of the plurality of generative models, the first improved version is obtained from each of the plurality of generative models. By inputting prompts to each of the multiple generative models, including a first selection request that requests the selection of the best first improved version from among several first improved versions, and a feedback request that requests the identification of missing information in the first improved version selected in response to the first selection request, the first selection information indicating the first improved version selected by each generative model and the missing information are obtained from each generative model. The selection unit selects one first improved version based on a plurality of first selection pieces of information. The dialogue unit inputs a prompt to the generation model that generated the first improved version selected by the selection unit, which includes an improvement request that requests further improvement of the first improved version selected by the selection unit by supplementing the missing information, thereby obtaining a second improved version which is a further improved version of the first improved version. The information processing system according to claim 1, wherein the display unit displays the second improved version.

6. It further comprises a selection section and a presentation section, The aforementioned dialogue unit, By inputting a prompt including the first improvement request to each of the plurality of generative models, the first improved version is obtained from each of the plurality of generative models. By inputting a prompt to each of the plurality of generative models, which includes a first selection request requesting the selection of the best first improved version from among the plurality of first improved versions, and a feedback request requesting the identification of missing information in the first improved version selected in response to the first selection request, the first selection information indicating the first improved version selected by each generative model and the missing information are obtained from each model. By inputting a prompt to each of the multiple generation models that includes a second improvement request requesting to further improve the first improved version selected by the selection unit by supplementing the missing information, multiple second improved versions, which are further improved versions of the first improved version, are obtained. By inputting a prompt to each of the multiple generative models that includes a second selection request requesting the selection of the best second improved version among the aforementioned second improved versions, second selection information indicating the second improved version selected by each generative model is obtained. The selection unit selects one of the second improved versions based on a plurality of the second selection pieces of information. The information processing system according to claim 3, wherein the presentation unit presents the second improved version selected by the selection unit.

7. An information processing system comprising a dialogue unit and a selection unit, The aforementioned dialogue unit, By inputting a prompt containing a generation request that requests the generation of predetermined information for each of the multiple generative models, responses are obtained from each of the multiple generative models. By inputting a prompt containing a selection request that asks for the selection of the best answer from among the aforementioned multiple answers into each of the multiple generative models, selection information indicating the best answer is obtained from each generative model. The selection unit selects one of the best answers based on the multiple selection pieces of information. The dialogue unit is an information processing system that obtains an improved version of the answer by inputting a prompt, which includes an improvement request, to the generation model that generated the answer selected by the selection unit, requesting an improvement to the answer based on the multiple answers.

8. It also includes a display section, The information processing system according to claim 7, wherein the display unit displays the improved version.

9. The aforementioned dialogue unit, By inputting a prompt to each of the multiple generative models, along with the multiple responses, which includes the selection request and the feedback request that points out missing information in the best response, the selection information and the missing information are obtained from each generative model. The information processing system according to claim 7, which obtains the improved version by inputting a prompt including the improvement request to the generation model that has generated the answer selected by the selection unit, which requests that the answer be improved by supplementing the missing information based on the plurality of answers.

10. A program that makes a computer function as an interactive unit, The aforementioned dialogue unit, By inputting a prompt containing a request to generate specific information for each of the multiple generative models, responses are obtained from each generative model. A program that obtains a first improved version of an answer by inputting a prompt to one generative model that includes a first improvement request requesting an improvement to an answer based on multiple answers obtained from the multiple generative models.

11. A program for causing a computer to function as an interactive unit and a selection unit, The aforementioned dialogue unit, By inputting a prompt containing a generation request that requests the generation of predetermined information for each of the multiple generative models, responses are obtained from each of the multiple generative models. By inputting a prompt containing a selection request that asks for the selection of the best answer from among the aforementioned multiple answers into each of the multiple generative models, selection information indicating the best answer is obtained from each generative model. The selection unit selects one of the best answers based on the multiple selection pieces of information. The dialogue unit is a program that obtains an improved version of the answer by inputting a prompt to the generation model that generated the answer selected by the selection unit, which includes an improvement request that requests an improvement to the answer based on the multiple answers.

12. An information processing method performed by a computer having a control unit, The control unit inputs a prompt containing a request for the generation of predetermined information to each of the multiple generative models, thereby obtaining a response from each generative model. The control unit inputs a prompt to one generation model that includes a first improvement request requesting improvement of the answer based on a plurality of answers obtained from the plurality of generation models, thereby obtaining a first improved version of the answer. Information processing methods, including those mentioned above.

13. An information processing method performed by a computer having a control unit, The control unit inputs a prompt to each of the multiple generative models, which includes a generation request that requests the generation of predetermined information, thereby obtaining a response from each of the multiple generative models. The control unit inputs a prompt to each of the multiple generative models, which includes a selection request that requests the selection of the best answer from among the multiple answers, thereby obtaining selection information indicating the best answer from each generative model. The control unit performs the step of selecting one of the best answers based on a plurality of selection pieces of information, The control unit inputs a prompt to the generation model that generated the selected answer, which includes an improvement request that requests an improvement to the answer based on the multiple answers, thereby obtaining an improved version of the answer. Information processing methods, including those mentioned above.

Citation Information

Patent Citations

  • Server, and computer program

    JP2024168208A