Information reasoning method, system, electronic device, storage medium, and program product

Abstract

The present disclosure relates to the fields of large model technology and natural language processing, and discloses an information reasoning method, a system, an electronic device, a storage medium, and a program product. The method comprises: detecting query information to undergo reasoning; performing decomposition processing on the query information, to obtain sub-query information at different reasoning stages, the sub-query information at different reasoning stages having a logical association relationship; determining sub-response information matching the sub-query information at different reasoning stages, respectively; verifying the sub-response information matching the sub-query information, to obtain a verification result corresponding to the sub-query information, the verification result being used to characterize a degree of feasibility of the sub-query information for reasoning query information; in response to the verification result being greater than a verification result threshold value, performing reasoning on the sub-query information corresponding to the verification result, to obtain reply information corresponding to the query information. The present disclosure solves the technical problem of low information reasoning efficiency.

Description

Information reasoning method, system, electronic device, storage medium and program product

[0001] This disclosure claims priority to the Chinese patent application filed with the China Patent Office on December 27, 2023, with application number 202311839894.9 and application name “Information Reasoning Method, System and Electronic Device,” the entire contents of which are incorporated by reference into this disclosure. Technical Field

[0002] The present disclosure relates to the fields of large model technology and natural language processing, and more specifically, to an information reasoning method, system, electronic device, storage medium, and program product. Background Art

[0003] In related technologies, in complex application scenarios such as logical reasoning and solving difficult mathematical problems, especially those involving many abstract concepts or multi-step reasoning, current information reasoning lacks certain cognitive capabilities. Solving complex logical reasoning or mathematical problems remains difficult and time-consuming, thus still presenting the technical problem of low information reasoning efficiency.

[0004] Currently, no effective solutions have been proposed for the above technical problems.

[0005] Summary of the Invention

[0006] The embodiments of the present disclosure provide an information reasoning method, system, electronic device, storage medium, and program product to at least solve the technical problem of low efficiency of information reasoning.

[0007] According to one aspect of an embodiment of the present disclosure, an information reasoning method is provided. The method may include the following steps: detecting query information to be reasoned; decomposing the query information to obtain sub-query information at different reasoning stages, wherein the sub-query information at different reasoning stages has a logical association relationship; determining sub-answer information that matches the sub-query information at different reasoning stages; verifying the sub-answer information that matches the sub-query information to obtain a verification result corresponding to the sub-query information, wherein the verification result is used to indicate the feasibility of the sub-query information for reasoning the query information; and, in response to the verification result being greater than a verification result threshold, reasoning on the sub-query information corresponding to the verification result to obtain answer information corresponding to the query information.

[0008] According to another aspect of an embodiment of the present disclosure, another information reasoning method is provided. The method may include the following steps: in response to query information to be reasoned received in a dialogue interface, displaying sub-query information of the query information at different reasoning stages on the dialogue interface, wherein the sub-query information at different reasoning stages has a logical association relationship; displaying sub-reply information matching the sub-query information at different reasoning stages on the dialogue interface; and displaying reply information corresponding to the query information on the dialogue interface, wherein the reply information is obtained by reasoning the sub-query information corresponding to the verification result when the verification result corresponding to the sub-query information is greater than a verification result threshold, and the verification result is obtained by verifying the sub-reply information matching the sub-query information, and is used to represent the feasibility of the sub-query information for the reasoned query information.

[0009] According to another aspect of an embodiment of the present disclosure, a model generation method is also provided. The method may include: obtaining a query information sample; using a sub-query information sample corresponding to the query information sample to train a first large model, wherein the first large model is used to decompose the input query information to obtain sub-query information at different inference stages, and the sub-query information at different inference stages has a logical association relationship; obtaining a reply information sample corresponding to the query information sample; obtaining a verification result sample corresponding to the reply information sample, wherein the verification result sample is used to at least represent a classification result of the reply information sample; using the verification result sample to train a second large model, wherein the second large model is used to verify the sub-reply information matched by the sub-query information to obtain a verification result corresponding to the sub-query information, the verification result being used to characterize the feasibility of the sub-query information for inferring the query information, and the sub-query information corresponding to the verification result greater than a verification result threshold is used to infer the reply information corresponding to the query information.

[0010] According to another aspect of an embodiment of the present disclosure, an information reasoning system is also provided. The system may include: an information generation end, configured to detect query information to be reasoned, decompose the query information to obtain sub-query information at different reasoning stages, and determine sub-response information that matches the sub-query information at different reasoning stages, wherein the sub-query information at different reasoning stages has a logical association relationship; and an information verification end, configured to verify the sub-response information that matches the sub-query information, obtain a verification result corresponding to the sub-query information, and, in response to the verification result being greater than a verification result threshold, perform reasoning on the sub-query information corresponding to the verification result to obtain reply information corresponding to the query information, wherein the verification result is used to indicate the feasibility of the sub-query information for reasoning the query information.

[0011] According to another aspect of an embodiment of the present disclosure, an electronic device is provided. The electronic device may include a memory and a processor: the memory is configured to store computer-executable instructions, and the processor is configured to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, any one of the aforementioned information reasoning methods is implemented.

[0012] According to another aspect of an embodiment of the present disclosure, a processor is provided, wherein the processor is configured to run a program, wherein any one of the above-mentioned information reasoning methods is executed when the program is running.

[0013] According to another aspect of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the storage medium is located is controlled to execute any one of the above-mentioned information reasoning methods.

[0014] According to another aspect of an embodiment of the present disclosure, a computer program product is further provided, including a computer program, which implements any of the above-mentioned information reasoning methods when executed by a processor.

[0015] In the disclosed embodiments, after detecting query information, the query information can be decomposed to determine sub-queries with logical relationships between them at different reasoning stages. Each sub-query can then be responded to, resulting in a corresponding sub-response. To ensure the feasibility of the determined responses, each sub-response can be verified, with each corresponding verification result. The feasibility of the query is determined by determining whether the verification result is greater than a verification result threshold. If the verification result is greater than the verification result threshold, the query information is considered highly feasible, and reasoning can be performed on the sub-queries to determine the response information. Considering the low efficiency of reasoning logical problems with logical relationships in related technologies due to the lack of cognitive capabilities of large models, the disclosed embodiments can divide the original query information into multiple sub-queries with logical relationships, based on the reasoning stages. By responding to the sub-queries at different reasoning stages, the response information can be ultimately determined, thereby significantly improving the accuracy of solving logical reasoning problems. This improves the efficiency of information reasoning and resolves the technical issue of low information reasoning efficiency.

[0016] It is easy to note that the above general description and the following detailed description are only for the purpose of exemplifying and explaining the present disclosure, and do not constitute a limitation of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The exemplary embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation of the present disclosure. In the drawings:

[0018] FIG1 is a schematic diagram of an application scenario of an information reasoning method according to an embodiment of the present disclosure;

[0019] FIG2 is a flow chart of an information reasoning method according to an embodiment of the present disclosure;

[0020] FIG3 is a flow chart of another information reasoning method according to an embodiment of the present disclosure;

[0021] FIG4 is a flow chart of a method for generating a model according to an embodiment of the present disclosure;

[0022] FIG5 is a schematic diagram of an information reasoning system according to an embodiment of the present disclosure;

[0023] FIG6 is a schematic diagram of a cognitive tree according to an embodiment of the present disclosure;

[0024] FIG7 is a schematic diagram of a complex task reasoning process according to an embodiment of the present disclosure;

[0025] FIG8 is a schematic diagram of an information reasoning device according to an embodiment of the present disclosure;

[0026] FIG9 is a schematic diagram of another information reasoning device according to an embodiment of the present disclosure;

[0027] FIG10 is a schematic diagram of a device for generating a model according to an embodiment of the present disclosure;

[0028] FIG11 is a structural block diagram of a computer terminal according to an embodiment of the present disclosure;

[0029] FIG12 is a block diagram of an electronic device according to an information inference method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the embodiments described are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present disclosure.

[0031] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0032] The technical solution provided by the present disclosure is mainly implemented using large-scale model technology. The large model here refers to a deep learning model with large-scale model parameters, which can usually contain hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than ten trillion model parameters. The large model can also be called a cornerstone model / foundation model (Foundation Model). The large model is pre-trained by large-scale unlabeled corpus to produce a pre-trained model with more than 100 million parameters. This model can adapt to a wide range of downstream tasks and has good generalization capabilities, such as large-scale language models (LLMs) and multi-modal pre-training models.

[0033] It should be noted that when the large model is actually applied, the pre-trained model can be fine-tuned with a small number of samples, so that the large model can be applied to different tasks. For example, the large model can be widely used in natural language processing (NLP), computer vision, speech processing and other fields. Specifically, it can be applied to computer vision tasks such as visual question answering (VQA), image captioning (IC), and image generation. It can also be widely used in natural language processing tasks such as text-based sentiment classification, text summary generation, and machine translation. Therefore, the main application scenarios of the large model include but are not limited to digital assistants, intelligent robots, search, online education, office software, e-commerce, and intelligent design.

[0034] First, some nouns or terms that appear in the description of the embodiments of the present disclosure are subject to the following explanations:

[0035] Large language model: A language model that can output corresponding response text based on natural language input.

[0036] Example 1

[0037] According to an embodiment of the present disclosure, an information reasoning method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0038] Taking into account the huge amount of model parameters of the large model and the limited computing resources of the mobile terminal, the above-mentioned information reasoning method provided by the embodiment of the present disclosure can be applied to the application scenario shown in Figure 1, but is not limited to this. In the application scenario shown in Figure 1, the large model is deployed in the server 10, which can be a cloud-based server. The server 10 can be connected to one or more client devices 20 via a local area network connection, a wide area network connection, an Internet connection, or other types of data networks. The client devices 20 here can include but are not limited to: smart phones, tablet computers, laptops, PDAs, personal computers, smart home devices, car-mounted devices, etc. The client devices together constitute the client relative to the server. An interactive interface for obtaining application generation instructions can be deployed on the graphical user interface of the client device, and the interactive interface can be a dialogue interface. The client device 20 can interact with the user through the graphical user interface to implement the call to the large model, thereby implementing the information reasoning method provided by the embodiment of the present disclosure.

[0039] In an embodiment of the present disclosure, a system consisting of a client device and a server can perform the following steps: If a client needs to query a large model to answer a question, it can enter query information corresponding to the question to be answered on the interactive interface of its client device. The client device can collect the query information and send it to the server via a network. After receiving the query information, the server can perform the following steps: Step S102: Detect the query information to be inferred; Step S104: Decompose the query information to obtain sub-query information at different inference stages, where the sub-query information at different inference stages has a logical relationship with each other; Step S106: Determine sub-answer information that matches the sub-query information at different inference stages; Step S108: Verify the sub-answer information that matches the sub-query information to obtain a verification result corresponding to the sub-query information, where the verification result is used to indicate the feasibility of the sub-query information for inferring the query information; Step S110: In response to the verification result being greater than a verification result threshold, reasoning is performed on the sub-query information corresponding to the verification result to obtain answer information corresponding to the query information. The answer information can be output to the client device.

[0040] During the above process, the sub-query information determined by the server, the sub-response information corresponding to the sub-query information, and the verification results obtained by verification can also be sent to the corresponding client device via the network. The sub-query information, sub-response information, and verification results obtained during the information reasoning process can be displayed on the client device's interactive interface. The interactive interface can be adjusted based on individual needs and accuracy to ensure the accuracy of the final required response information. After the server determines the response information, it can be transmitted to the corresponding client device via the network. After the client device receives the response information, the response information can be displayed on the interactive interface.

[0041] Considering that the efficiency of reasoning logical problems with logical relationships is low in related technologies due to the lack of cognitive ability of large models, the embodiments of the present disclosure can divide the original query information into multiple sub-query information with logical associations according to the before and after reasoning stages. By responding to the sub-query information at different reasoning stages, the response information of the query information can be finally determined, thereby greatly improving the accuracy of the large model in solving logical reasoning and other problems, thereby achieving the technical effect of improving the efficiency of information reasoning and solving the technical problem of low efficiency of information reasoning. It should be noted that the embodiments of the present disclosure can be carried out in the client device if the operating resources of the client device can meet the deployment and operation conditions of the large model.

[0042] The embodiment of the present disclosure proposes the following method from the technical implementation side. Under the above operating environment, the embodiment 1 of the present disclosure provides an information reasoning method as shown in Figure 2. Figure 2 is a flow chart of an information reasoning method according to the present disclosure. As shown in Figure 2, the method may include the following steps:

[0043] Step S202: Detect query information to be inferred.

[0044] In the technical solution provided in the above step S202 of the present disclosure, when the client has a question that needs to be answered by querying the large model, the question to be answered can be constituted as query information and input into the interactive interface of the client. When the query information is detected to be input on the interactive interface, the query information collected on the interactive interface can be transmitted to the server through the network. The server can detect whether the query information to be inferred is received, wherein the query information can be the original question (original question) that the user of the client needs to be answered by the large model, which can also be called the initial query. The original question can be the query question (Q) input by the user into the interactive interface. The query Q can be a question that requires the large model to use its own cognitive ability to reason to answer, for example, it can be a more complex task such as a mathematical problem (Math Problem) or a logical reasoning problem. It should be noted that the above query information is only for example and is not specifically limited here. As long as it is a problem that requires the large model to use cognitive ability to reason and analyze to answer, it is within the protection scope of the embodiment of the present disclosure.

[0045] Optionally, when a user on the client has a question that needs to be asked to the big model, query information that can describe the question being queried can be entered on the interactive interface of the user's client device. For example, natural language that can describe the more complex logical reasoning problem can be entered. The query information input can be confirmed by clicking the corresponding control on the interactive interface. At this time, the interactive interface can send the collected query information to the server through the network. Since the server is deployed with a big model that can reason about the query information. Therefore, the big model can be used to process the query information in the server. It should be noted that the above language to describe the query information is only an example and is not specifically limited here.

[0046] Optionally, before reasoning about the query information, a large model with strong cognitive ability, strong logical reasoning and analysis capabilities can be pre-trained and pre-deployed in the server to provide question-and-answer services for the clients associated with the server. That is, by deploying a server with a large model, users can be provided with logical reasoning and data problem-solving services.

[0047] Optionally, after the server receives the query information transmitted by the network, it can enable a pre-deployed large model to reason and analyze the query information.

[0048] Step S204 : Decompose the query information to obtain sub-query information at different reasoning stages, wherein the sub-query information at different reasoning stages has a logical association relationship.

[0049] In the technical solution provided in step S204 of the present disclosure, after detecting the query information to be inferred, the query information can be decomposed to obtain sub-query information at different inference stages, wherein the sub-query information at different inference stages has a certain logical relationship. The different inference stages can be used to represent the solution process for the query information. For example, if solving the query information requires two inference stages, it can include a first inference stage (Step 1) and a second inference stage (Step 2). The inference stage can be a chain of thought, a decomposition set, or a generated result obtained by analyzing the query information to solve the query information. The sub-query information can be used to represent the stage-by-stage analysis results obtained by reasoning in the corresponding inference stage. For example, it can be a decomposition hypothesis, decomposition D, or a smaller problem obtained by decomposition. Each inference stage can include at least one sub-query information. For example, Step 1 can include three sub-query information 1a, 1b, and 1c, and Step 2 can include three sub-query information 2a, 2b, and 2c. The sub-query information can be a sub-problem decomposed from the original problem, and solving the sub-problem helps solve the entire original problem. Logical association relationships can be used to represent continuous language sequences between sub-query information in the reasoning phase, for example, 1a in Step 1 and 2b in Step 2.

[0050] It should be noted that the number of reasoning stages and the number and specific content of sub-queries decomposed in different reasoning stages are merely illustrative and are not intended to be limiting. Any process or method capable of dividing a complex query into multiple, simpler sub-queries according to the stages of reasoning and analysis, and then resolving each sub-query to solve the entire query, falls within the scope of protection of the presently disclosed embodiments.

[0051] In related technologies, since large language models lack cognitive and logical reasoning capabilities, it is easier to handle simpler conversations. However, when faced with answering more complex mathematical or logical reasoning problems, not only does the reasoning take a long time, but the answers to the questions cannot be obtained. Therefore, there is still a technical problem of low efficiency in information reasoning.

[0052] As an optional example, in the embodiment of the present disclosure, taking into account the above-mentioned problems, a new reasoning framework can be designed for a large model to enhance the reasoning ability of the large model on complex mathematical problems, etc. Through this reasoning framework, an original relatively complex problem that is difficult to solve at one time can be decomposed and processed into multiple relatively simple small problems that can be solved at one time according to the reasoning stage, that is, the complex query information can be decomposed and processed into simple sub-query information at different reasoning stages. In this way, the entire complex original problem can be finally solved by gradually reasoning to solve small problems, that is, by answering the sub-query information, the reply information of the final query information is obtained. By decomposing and reasoning complex problems in the above-mentioned way, the cognitive ability of the large model and its reasoning ability for complex logical reasoning problems are enhanced, thereby achieving the technical effect of improving the efficiency of information reasoning.

[0053] Step S206: Determine sub-answer information that matches the sub-query information at different reasoning stages.

[0054] In the technical solution provided in the above step S206 of the present disclosure, after the query information is decomposed and processed to obtain sub-query information of different reasoning stages, the sub-response information corresponding to each sub-query information can be determined respectively, wherein the sub-response information can be used to represent the analysis result of the question corresponding to the sub-query information.

[0055] Optionally, the sub-query information contained in each reasoning stage obtained by decomposing the query information can be reasoned and analyzed one by one to determine the solution process and final result of each sub-query information, that is, the sub-response information corresponding to each sub-query information can be determined through reasoning and analysis.

[0056] Considering that it is very difficult for a large model to solve a relatively complex problem at one time. Therefore, in the embodiment of the present disclosure, in order to ensure the efficiency of problem solving and reduce the difficulty of problem solving, an intuitive system (Intuitive System) can be constructed in the reasoning framework of the large model, and the original problem can be analyzed and decomposed by the intuitive system to generate decomposed hypotheses, that is, to generate sub-query information in different reasoning stages. By solving simple small problems one by one, the difficulty of solving the complex original problem is weakened, thereby achieving the technical effect of improving the efficiency of large models in solving problems, wherein the generation result of the intuitive system is the sub-response information of the sub-query information.

[0057] Step S208 , verifying the sub-response information matched by the sub-query information to obtain a verification result corresponding to the sub-query information, wherein the verification result is used to represent the feasibility of the sub-query information for the inference query information.

[0058] In the technical solution provided in the above step S208 of the present disclosure, after respectively determining the sub-reply information that matches the sub-query information at different reasoning stages, the sub-reply information that matches the sub-query information can be verified to obtain a verification result corresponding to the sub-query information, wherein the verification result can be used to characterize the feasibility of the sub-query information to the reasoning query information, that is, it can be used to evaluate the feasibility and acceptability of the assumptions of the sub-reply information, and can be obtained by scoring the feasibility corresponding to the sub-reply information. The verification result can be a decomposition score (Score|Decomposition). For example, the verification result can be a feasibility score: 30 points, 60 points or 100 points, or it can be Sure, Likely or Impossible. It should be noted that the scoring form of the above verification results is only for illustration and is not specifically limited here. As long as it is a form and method that can score the sub-reply information corresponding to the sub-query information obtained by decomposing the query information, it is within the protection scope of the embodiments of the present disclosure.

[0059] Optionally, in order to ensure the accuracy of the reply information of the query information finally determined, it is necessary to evaluate the sub-reply information obtained by solving each sub-query information obtained by decomposition to evaluate the acceptability of the sub-reply information. This method can be used to ensure the accuracy of the determined sub-reply information, thereby ensuring the accuracy of the reply information of the final query information.

[0060] In the disclosed embodiment, in order to ensure the accuracy of solving complex logical reasoning problems, etc., a reflective system (Reflective System) can be constructed in the reasoning framework of the large model. The reflective system can be used to evaluate the sub-response information solved by the sub-query information decomposed and processed in the intuitive system. That is, the role of the reflective system is to evaluate the generation results of the intuitive system to determine its acceptability, and to obtain the verification results after the evaluation. Through the verification results, the accuracy of the sub-response information can be ensured, thereby ensuring the accuracy of the reply information, and thus achieving the technical effect of improving the accuracy of the large model reasoning problem.

[0061] Step S210 , in response to the verification result being greater than the verification result threshold, reasoning is performed on the sub-query information corresponding to the verification result to obtain reply information corresponding to the query information.

[0062] In the technical solution provided in the above step S210 of the present disclosure, after verifying the sub-response information that matches the sub-query information and obtaining the verification result corresponding to the sub-query information, the size relationship between the verification result and the verification result threshold can be determined. If the verification result is greater than the verification result threshold, it can be said that the sub-query information is more feasible for reasoning the query information, and the sub-query information corresponding to the verification result can be inferred to obtain the reply information of the query information, wherein the verification result can be used to indicate the correctness of the determined sub-query information and the sub-response information, that is, the reflective system verifies the correctness of the hypothesis obtained by the intuitive system. The reply information can be used to indicate the final result corresponding to the query information. The verification result threshold can be a pre-set result, or it can be a result set by itself based on the actual information reasoning situation. The verification result threshold can be 99 points or Sure.

[0063] It should be noted that the setting method and numerical value of the above-mentioned verification results are only for illustration and are not specifically limited here. As long as the sub-query information and sub-response information can be evaluated to ensure the accuracy of the reply information, the conditions set are within the protection scope of the embodiments of the present disclosure.

[0064] Optionally, after verifying the sub-query information and sub-reply information and obtaining a verification result, the magnitude relationship between the verification result and the verification result threshold can be determined. If the verification result is less than or equal to the verification result threshold, it can be indicated that the correctness of the obtained sub-query information and sub-reply information is low. If the reply information of the final query information is inferred based on the sub-reply information at this time, the accuracy cannot be guaranteed. If the verification result is greater than the verification result threshold, it can be indicated that the correctness of the obtained sub-query information and sub-reply information is high. In this case, the reply information of the final query information can be inferred based on the sub-reply information to ensure the accuracy of the reply information.

[0065] In the disclosed embodiments, the intuitive system and reflective system within a novel reasoning framework designed for a large model enhance the model's logical reasoning and cognitive capabilities. The intuitive system simplifies complex problems. Specifically, the reasoning phases of a complex problem can be divided according to logical relationships and reasoning capabilities, breaking the complex problem into multiple, simpler sub-problems. The intuitive system can then solve each of these simple sub-problems, improving the large model's ability to solve complex problems. However, simplifying complex problems solely through the intuitive system cannot guarantee the accuracy of the problem solution. Therefore, a reflective system can be designed to supplement this, verifying the sub-query and sub-response information analyzed by the intuitive system to ensure accurate solutions. In other words, the intuitive and reflective systems constructed for the large model use the intuitive system to generate hypotheses for decomposing the original problem, and the reflective system to verify the correctness of the hypotheses, guiding the subsequent generation of the intuitive system. Thus, in the disclosed embodiments, the intuitive and reflective systems complement each other to ensure the accuracy and efficiency of the large model's answers, thereby achieving the technical effect of improving the efficiency of information reasoning.

[0066] Through the above steps S202 to S210 of the present disclosure, after detecting the presence of query information in the interactive interface of a certain client, the query information can be transmitted from the client to the server, and the query information can be decomposed and processed in the server to determine the sub-query information that has a logical association with each other in different reasoning stages. And each sub-query information can be replied to and the corresponding sub-reply information can be obtained. In order to ensure the feasibility of the determined reply information, each sub-reply information can be verified, and each sub-reply information corresponds to a verification result. Whether the query information is feasible is determined by judging whether the verification result is greater than the verification result threshold. If the verification result is greater than the verification result threshold, it means that the query information is highly feasible, and the sub-query information can be reasoned to determine the reply information of the query information. Taking into account the low efficiency of reasoning logical problems with logical relationships in related technologies due to the lack of cognitive ability of large models, the embodiments of the present disclosure can divide the original query information into multiple sub-query information with logical associations according to the before and after reasoning stages. By responding to the sub-query information at different reasoning stages, the reply information of the query information can be finally determined, thereby greatly improving the accuracy of the large model in solving logical reasoning and other problems, thereby achieving the technical effect of improving the efficiency of information reasoning and solving the technical problem of low efficiency of information reasoning.

[0067] The above method of this embodiment is further introduced below.

[0068] As an optional implementation, step S204 decomposes the query information to obtain sub-query information of different reasoning stages, including: based on the query information, searching for sub-query information of different reasoning stages in the sample set, wherein the sample set includes sub-query information corresponding to different query information.

[0069] In this embodiment, in the process of decomposing the query information to obtain sub-query information of different reasoning stages, based on the query information, sub-query information of different reasoning stages can be found in the sample set, where the sample set can include sub-query information corresponding to different query information. The sample set can also be called a training set, a decomposition set, a context example (in-context learning) or a context, etc.

[0070] Optionally, before using a large model for information reasoning, a sample set can be prepared in advance, in which a large amount of query information can be summarized, and the sub-query information that can be obtained by dividing each query information according to the reasoning stage can be summarized for the server to call.

[0071] Optionally, a pre-defined sample set can be deployed in the intuitive system. Thus, after the server detects query information transmitted by a client device, it can traverse the query information from the sample set in the intuitive system and find the sub-query information corresponding to the query information.

[0072] Optionally, in the case of logical reasoning problems, the sub-query information corresponding to the query information can be determined from the sample set through the intuitive system. The above method can be used to decompose more complex problems into smaller problems. By reasoning about these decomposed small problems, the ultimate goal can be achieved, that is, the response information of the query information can be obtained.

[0073] As an optional implementation, based on the query information, sub-query information of different reasoning stages is searched in the sample set, including: determining the identification information of the query information; in the sample set corresponding to the reasoning stage, the sub-query information whose identification information matches the identification information of the query information is determined as the sub-query information of the reasoning stage.

[0074] In this embodiment, in the process of searching for sub-query information of different inference stages in the sample set, the identification information of the query information can be determined. According to the identification information of the query information sent by the current client device, it can be determined from the sample set whether there is identification information corresponding to the identification information. If so, the sub-query information matched by the identification information can be determined as the sub-query information of the current query information in the inference stage, wherein the identification information can be used to represent the current query.

[0075] Optionally, after receiving the query information sent by the client, the server can decompose, process, and match the query information based on the intuitive system in the server's large model. The generative capabilities of the intuitive system are the foundation for building the cognitive tree. The capabilities of the intuitive system can be enhanced through contextual methods. The query Q can be defined as the ultimate goal of a logical reasoning problem or a data problem, that is, the query information.

[0076] For example, the intuitive system can select a decoder-only model, such as the Generative Pre-trained Transformer (GPT) 2-XL model or the Language Learning and Multilingualism Alliance (LLaMA)-7B model, as the intuitive system. It should be noted that the models used in the above intuitive systems are only examples and are not specifically limited here. As long as the model can decompose and match the query information, it is within the scope of protection of the embodiments of the present disclosure.

[0077] Alternatively, in the case of a logical reasoning problem, decomposition D involves further decomposing the query information into smaller questions, and the final answer information can be determined by reasoning on the above decomposition.

[0078] Optionally, in the case of a mathematical problem, decomposition D refers to subproblems decomposed from the original problem, and the final answer information is determined through the decomposed subproblems, that is, solving the subproblems helps to solve the entire original problem.

[0079] Optionally, the decomposition set may represent a decomposition set of examples in the training set, that is, the sub-query information contained in the sample set is obtained by decomposing a large amount of query information in advance.

[0080] Optionally, k examples can be retrieved from the inference decomposition set (e.g., query: Q; decomposition: query D), and then the above examples can be used as context for model input. The output can be generated as y~f θ (y|x,z 1…k ), where [y]~f θ (y|x,z 1…k ) can be used to represent a continuous language sequence. z can be used to represent the k examples retrieved from the decomposition set Z, Z = {z1,…z L}.

[0081] As an optional implementation, in the sample set corresponding to the reasoning stage, the sub-query information that matches the identification information with the identification information of the query information is determined as the sub-query information of the reasoning stage, including: in the sample set corresponding to the reasoning stage, retrieving the sub-query information whose similarity between the identification information and the identification information of the query information is greater than the similarity threshold; and determining the retrieved sub-query information as the sub-query information of the reasoning stage.

[0082] In this embodiment, in the process of determining sub-query information whose identification information matches the identification information of the query information in the sample set corresponding to the inference stage as the sub-query information for the inference stage, sub-query information whose similarity between the identification information and the identification information of the query information is greater than a similarity threshold can be retrieved from the sample set corresponding to the inference stage, and such sub-query information can be determined as the sub-query information for the inference stage. The similarity can be used to represent the degree of similarity between the query information in the sample set and the query information issued by the client, and the similarity can be cosine similarity. It should be noted that the cosine similarity mentioned above is for illustration only and is not a specific limitation herein.

[0083] Optionally, the similarity threshold can be a value set according to the actual information reasoning, or a pre-set value, for example, the similarity threshold can be pre-set to 99.9%. It should be noted that the setting method and value of the similarity threshold are only examples and are not specifically limited here.

[0084] Optionally, after obtaining the query information sent by the current client, the query information in the sample set can be traversed and the similarity between each query information and the query information sent by the current client can be determined. If the similarity reaches a similarity threshold, it can be considered that the two are highly similar. In this case, the query information in the sample set that is highly similar to the query information sent by the current client, as well as the sub-query information corresponding to the query information, can be used as the corresponding sub-query information of the query information sent by the current client in the inference phase.

[0085] Optionally, an intuitive system can be used to obtain a representation of the current query. Specifically, the intuitive system can be used to obtain the identification information of the query information currently issued by the client. Cosine similarity can be calculated with representations of other queries in the set. Specifically, the identification information of the query information included in the sample set can be traversed. During the traversal process, the cosine similarity between each identification information in the sample set and the identification information of the query information issued by the client can be determined. K queries that meet the requirements can be retrieved from the set. Specifically, at least one query information with a cosine similarity greater than a similarity threshold can be retrieved.

[0086] As an optional implementation manner, determining the identification information of the query information includes: analyzing the query information using the first large model to obtain the identification information of the query information.

[0087] In this embodiment, in the process of determining the identification information of the query information, the first large model can be used to analyze the query information to determine the identification information of the query information, wherein the first large model can be an intuitive system, or it can be called an implicit extraction module, which is a smaller-scale model.

[0088] Optionally, when a client requests a large model to answer a query, the user can enter query information describing the query content on the client's interactive interface. The query information collected by the interactive interface can be transmitted to the server via the network. The intuitive system in the server can analyze the query information of the current client to obtain identification information of the query information. In other words, the intuitive system can obtain a representation of the current query.

[0089] As an optional implementation, in the sample set corresponding to the inference stage, the sub-query information whose identification information matches the identification information of the query information is determined as the sub-query information of the inference stage, including: using the first largest model to analyze the sample set corresponding to the inference stage and the identification information of the query information to obtain the sub-query information whose identification information matches the identification information of the query information; and using the first largest model to determine the matched sub-query information as the sub-query information of the inference stage.

[0090] In this embodiment, in the process of determining sub-query information whose identification information matches the identification information of the query information in the sample set corresponding to the inference stage as the sub-query information for the inference stage, a first large model can be used to analyze the sample set corresponding to the inference stage and the identification information of the query information, thereby determining the sub-query information corresponding to the identification information that matches the identification information of the current query information. The first large model can also be used to determine the matched sub-query information as the sub-query information for the inference stage, where the first large model can be a model corresponding to the intuitive system.

[0091] Optionally, after using the intuitive system to obtain the identification information corresponding to the query information issued by the current client, the intuitive system can also be used to call a sample set stored in the intuitive system based on the current identification information. The intuitive system traverses the identification information of each query information in the sample set and determines the similarity between each identification information and the identification information of the current query information. It can be determined whether the similarity exceeds a similarity threshold. If so, the sub-query information corresponding to the identification information in the sample set can be determined as the sub-query information of the current client's query information during the inference phase.

[0092] Alternatively, supervised fine-turning (SFT) has been shown to be effective in recognizing user intent. Therefore, in the disclosed embodiment, the intuitive system can use contextual examples to decompose the query Q into sub-questions. That is, the intuitive system can decompose the query information samples of a complex question into corresponding sub-query information samples. Since a generative model can be used as the intuitive system, during the training of the intuitive system, the loss of the sub-query information samples generated based on the query information samples can be calculated to determine the loss degree of the sub-query information. The loss degree and the sample set can then be used to train the intuitive system.

[0093] Optionally, during the autoregressive calculation, the generated text without the given context can be used, that is, the loss is calculated only for the subquery information sample. For example, given a sample of length N, it can be represented by X, where X = {x1, ...x i ,…x n}. The sequence length of the context example can be defined as M. The standard language modeling objective is used to maximize the following likelihood function: This trains the intuitive system.

[0094] As an optional implementation, determining sub-response information that matches sub-query information at different reasoning stages includes: using the first largest model to analyze sub-query information at different reasoning stages to obtain sub-response information that matches sub-query information at different reasoning stages.

[0095] In this embodiment, in the process of determining the sub-response information that matches the sub-query information of different reasoning stages, the first large model can be used to analyze the sub-query information of different reasoning stages to obtain the sub-response information that matches the sub-query information of different reasoning stages.

[0096] Optionally, after using the first model to determine the identification information of the query information that matches the identification information of the current query information from the sample set, the sub-query information corresponding to the identification information can be determined. The first large model can be used to solve each sub-query information separately to obtain the sub-answer information corresponding to each sub-query information.

[0097] Optionally, after using the intuitive system to analyze the current query information and determine the sub-query information of the query information in the reasoning stage, the sub-query information under different reasoning stages can be analyzed to solve the sub-response information corresponding to each sub-query information, so as to finally solve the reply information of the query information.

[0098] In the embodiment of the present disclosure, when faced with problems that require logical reasoning or solving more complex mathematical problems, the process of human cognition can be imitated through the intuitive system and the reflective system based on human cognitive theory, that is, the above two systems are used to improve cognitive ability. The intuitive system is responsible for generating multiple decomposition hypotheses of the original problem and solving the multiple decomposition hypotheses one by one. The reflective system verifies the hypotheses and solutions generated by the intuitive system, and selects the more likely hypotheses for subsequent generation until the final result is reached. Through the iterative generation of the above dual systems, the purpose of improving the problem-solving accuracy of the large model can be achieved, thereby achieving the technical effect of improving the accuracy of information reasoning of the large model.

[0099] As an optional implementation, the sub-reply information that matches the sub-query information is verified to obtain a verification result corresponding to the sub-query information, including: when running to the current reasoning stage in different reasoning stages, the sub-reply information that matches the sub-query information in the current reasoning stage is verified to obtain a sub-verification result of the current reasoning stage, wherein the verification result includes the sub-verification result; or, the sub-reply information in different reasoning stages is verified to obtain an overall verification result of the different reasoning stages, wherein the verification result includes the overall verification result.

[0100] In this embodiment, in the process of verifying the sub-response information matched by the sub-query information and obtaining the verification result corresponding to the sub-query information, when running to each reasoning stage in different reasoning stages, the sub-response information matched by the sub-query information of the current reasoning stage can be verified to obtain the sub-verification result of the current reasoning stage, or the sub-response information of different reasoning stages can be verified as a whole to obtain the overall verification result of different reasoning stages, wherein the verification result can include the sub-verification result and the overall verification result. The sub-verification result can be used to represent the verification of the intermediate process, and can be the score of the current state. The overall verification result can be used to represent the verification of the entire reasoning chain, and can be the overall score of the entire reasoning chain.

[0101] In the disclosed embodiment, the sub-response information generated by the first large model can be evaluated to determine its acceptability. Two methods can be used to verify this information and obtain verification results: verification of the intermediate process and verification of the entire reasoning chain. This ensures the rationality of the generated hypotheses and the reasoning process, thereby achieving the technical effect of improving the accuracy of information reasoning performed by the large model.

[0102] As an optional implementation method, the sub-reply information of the current reasoning stage in different reasoning stages is verified to obtain the sub-verification result of the current reasoning stage, including: using the second largest model to verify the sub-reply information of the current reasoning stage to obtain the sub-verification result; verifying the sub-reply information of different reasoning stages to obtain the overall verification result of different reasoning stages, including: using the second largest model to verify the sub-reply information of different reasoning stages to obtain the overall verification result.

[0103] In this embodiment, the second largest model can be used to verify the sub-response information of the current reasoning stage to obtain a sub-verification result, or the second largest model can be used to verify the sub-response information of different reasoning stages to obtain an overall verification result, wherein the second largest model can be a reflective system, or it can be called an explicit reasoning module.

[0104] Optionally, after using the intuitive system to decompose the query information and obtain sub-query information and sub-response information corresponding to the sub-query information, the sub-response information can be transmitted to the reflective system, and the reflective system can be used to evaluate the acceptability of the sub-response information determined by the intuitive system.

[0105] Alternatively, the reflective system functions differently from the intuitive system. While the intuitive system relies on quick intuition for generation, the reflective system's role is to evaluate the intuitive system's output to determine its acceptability. The reflective system can verify its output in two ways: by verifying intermediate processes and by verifying the overall chain of reasoning.

[0106] Optionally, based on the above analysis, a second largest model with the same model architecture as the first largest model can be determined, and the second largest model can be used to verify the sub-response information in different reasoning stages to obtain verification results.

[0107] For example, for the verification of the intermediate process, given the current state S (query: Q and decomposition: D), a reflective system with the same model architecture as the intuitive system can be used to generate a score v for verifying the current state, where v can be expressed as V(f θ ,s)~f θ (v|s).

[0108] For another example, for the verification of the whole reasoning chain, the complete reasoning chain can be given as S = {s1,…,s i ,…,s n}, a reflection system can be used to produce an overall fraction o, where o can be expressed as O(f θ ,S)~f θ (o|S).

[0109] In the disclosed embodiment, the reflective system is different from the intuitive system in that its important task is to evaluate and verify the feasibility of the entire reasoning chain of the current state, rather than generating quick hypotheses like the intuitive system. This evaluation process helps to determine whether the generated hypotheses and reasoning process are reasonable, thereby achieving the technical effect of improving the accuracy of information reasoning of large models.

[0110] Optionally, in the process of training the reflective system, the same training method as the intuitive system can be adopted, using positive and negative samples to allow the generative model to generate classification results from the positive and negative samples, that is, to generate corresponding verification result samples, by determining the loss degree of the verification result samples.

[0111] Alternatively, since the reflection system mainly focuses on the judgment of state s, the following loss function can be used to determine the loss of the verification result sample:

[0112] As an optional implementation method, the query information is decomposed and processed to obtain sub-query information of different reasoning stages, including: searching for sub-query information of the current reasoning stage based on the query information in the sample set corresponding to the current reasoning stage; when the different reasoning stages include the next reasoning stage of the current reasoning stage, searching for sub-query information of the next reasoning stage based on the sub-verification result and query information of the current reasoning stage in the sample set corresponding to the next reasoning stage.

[0113] In this embodiment, sub-query information for the current reasoning stage can be searched for in the sample set corresponding to the current reasoning stage based on the query information. If different reasoning stages include the next reasoning stage of the current reasoning stage, sub-query information for the next reasoning stage can be searched for in the sample set corresponding to the next reasoning stage based on the sub-verification result of the current reasoning stage and the query information.

[0114] In this disclosed embodiment, two main components are constructed: an intuitive system and a reflective system. The intuitive system can quickly generate multiple answers using contextual examples, while the reflective system can use comparative learning to score the intuitive system's answers. This score can guide the intuitive system in subsequent generation steps.

[0115] Optionally, after the reflective system determines the verification result, the verification result can be sent to the intuitive system. The intuitive system can continue with the subsequent generation steps based on the reflective system's evaluation. That is, after the reflective system evaluates the sub-query information of a certain reasoning stage and obtains the sub-verification result, it can send the sub-verification result to the intuitive system. The intuitive system can reason about the sub-query information in the next reasoning stage of the current reasoning stage based on the sub-verification result. If all reasoning stages are reasoned and evaluated, the entire reasoning chain can be evaluated by the reflective system, that is, the overall verification result can be determined.

[0116] As an optional implementation, the method may also include: determining the query information as the root node, determining the sub-query information as the leaf node, and establishing a target tree structure; reasoning on the sub-query information corresponding to the verification result to obtain reply information corresponding to the query information, including: in the target tree structure, reasoning on the sub-query information corresponding to the verification result to obtain reply information corresponding to the query information.

[0117] In this embodiment, the query information can be determined as the root node of a tree, and the sub-query information can be determined as the leaf nodes of the tree, thereby establishing a target tree structure. In the target tree structure, the sub-query information corresponding to the verification result can be inferred to obtain the response information corresponding to the query information. The target tree structure can be a cognitive tree (CogTree), also known as an inference tree.

[0118] Optionally, the reasoning framework can adopt an iterative method to build a cognitive tree structure based on human thinking patterns. The root node of the cognitive tree can represent the initial query, that is, the query information, and the leaf nodes can contain simple questions that are directly answered, that is, the sub-query information.

[0119] Optionally, in order to lightweight complex task reasoning, a smaller-scale model can be used to build a dual-system generative reasoning tree to effectively enhance the large model's ability to answer complex mathematical and logical reasoning problems.

[0120] Optionally, the cognitive tree uses a smaller and more open-source model to solve the problem of high model inference cost, and the inference tree is a process of generating an inference tree using a dual system to enhance the problem of insufficient inference ability of the large model itself.

[0121] In the disclosed embodiment, the large model can construct two systems: an intuitive system and a reflective system. The intuitive system can be used to generate hypotheses for decomposing the original problem, and the reflective system can be used to verify the correctness of the hypotheses and guide subsequent generation of the intuitive system. By iteratively generating an inference tree using the two systems, the reasoning ability of the large model is enhanced, thereby enhancing the large model's reasoning ability on complex mathematical and logical reasoning problems, thereby achieving the technical effect of improving the accuracy of the large model's information reasoning.

[0122] The embodiment of the present disclosure also provides an information reasoning method on the human-computer interaction side. Figure 3 is a flow chart of an information reasoning method according to the embodiment of the present disclosure. As shown in Figure 3, the method may include the following steps:

[0123] Step S302 : In response to the query information to be inferred received in the dialogue interface, sub-query information of the query information at different inference stages is displayed on the dialogue interface, wherein the sub-query information at different inference stages has a logical association relationship.

[0124] In the technical solution provided in the above step S302 of the present disclosure, in response to the query information to be inferred received in the dialogue interface, sub-query information of the query information at different reasoning stages can be displayed on the dialogue interface, wherein the sub-query information at different reasoning stages has a logical association relationship.

[0125] Optionally, if the client needs to query the large model to answer a question, the client can enter the query information corresponding to the question to be answered on the interactive interface on the client device, and the query information can be displayed on the dialogue interface.

[0126] Alternatively, the client device may collect the query information and send it to the server via the network. Since the server is deployed with a large model capable of reasoning about the query information, the server can use the large model to process the query information.

[0127] Optionally, the server may decompose the query information to obtain sub-query information at different inference stages, and may send the inferred sub-query information to the client device and display it on the dialogue interface.

[0128] Optionally, if the user believes that the decomposed sub-query information does not meet the requirements, the user may adjust it through corresponding operations.

[0129] Step S304: displaying sub-answer information that matches the sub-query information at different reasoning stages on the dialogue interface.

[0130] In the technical solution provided in the above step S304 of the present disclosure, sub-answer information respectively matching the sub-query information at different reasoning stages may also be displayed on the dialogue interface.

[0131] Optionally, the sub-query information contained in each reasoning stage obtained by decomposing the query information can be reasoned and analyzed one by one to determine the solution process and final result of each sub-query information, that is, the sub-response information corresponding to each sub-query information can be determined through reasoning and analysis.

[0132] Optionally, the sub-answer information corresponding to the sub-query information solved by the server can be sent to the corresponding client device and displayed on the interactive interface of the client device. If the user believes that the solved sub-answer information does not meet the requirements, he can adjust it through corresponding operations.

[0133] Step S306: Display the reply information corresponding to the query information on the dialogue interface, wherein the reply information is obtained by inferring the sub-query information corresponding to the verification result when the verification result corresponding to the sub-query information is greater than the verification result threshold, and the verification result is obtained by verifying the sub-reply information matching the sub-query information, and is used to characterize the feasibility of the sub-query information for the inferred query information.

[0134] In the technical solution provided in step S306 of the present disclosure, the response information corresponding to the query information can be displayed on the dialogue interface. The response information can be derived by inferring the sub-query information corresponding to the verification result, when the verification result corresponding to the sub-query information is greater than a verification result threshold. The verification result can be derived by verifying the sub-response information that matches the sub-query information and can be used to indicate the feasibility of the sub-query information for the inferred query information.

[0135] Optionally, the sub-query information contained in each reasoning stage obtained by decomposing the query information can be reasoned and analyzed one by one to determine the solution process and final result of each sub-query information, that is, the sub-response information corresponding to each sub-query information can be determined through reasoning and analysis.

[0136] Optionally, after respectively determining the sub-answer information that matches the sub-query information at different reasoning stages, the sub-answer information that matches the sub-query information may be verified to obtain a verification result corresponding to the sub-query information.

[0137] Optionally, in order to ensure the accuracy of the reply information of the query information finally determined, it is necessary to evaluate the sub-reply information obtained by solving each sub-query information obtained by decomposition to evaluate the acceptability of the sub-reply information. This method can be used to ensure the accuracy of the determined sub-reply information, thereby ensuring the accuracy of the reply information of the final query information.

[0138] Optionally, after verifying the sub-reply information that matches the sub-query information and obtaining the verification result corresponding to the sub-query information, the size relationship between the verification result and the verification result threshold can be judged. If the verification result is greater than the verification result threshold, it can be said that the sub-query information is more feasible for inferring the query information, and the sub-query information corresponding to the verification result can be inferred to obtain the reply information of the query information.

[0139] Optionally, if the verification result is less than or equal to the verification result threshold, it can be indicated that the correctness of the obtained sub-query information and sub-answer information is low. If the answer information of the final query information is inferred based on the sub-answer information at this time, the accuracy cannot be guaranteed. If the verification result is greater than the verification result threshold, it can be indicated that the correctness of the obtained sub-query information and sub-answer information is high. In this case, the answer information of the final query information can be inferred based on the sub-answer information to ensure the accuracy of the answer information.

[0140] Optionally, after the reply information is determined, the reply information may be transmitted to a corresponding client device via a network and may be displayed on a dialogue interface of the client device.

[0141] Optionally, if the user believes that the requested answer information does not meet the requirements, the user may adjust it through corresponding operations.

[0142] Through the above-mentioned steps S302 to S306 of the present invention, in response to the query information to be inferred received in the dialogue interface, sub-query information of the query information at different reasoning stages is displayed on the dialogue interface, wherein the sub-query information at different reasoning stages has a logical association relationship; sub-reply information matching the sub-query information at different reasoning stages is displayed on the dialogue interface; reply information corresponding to the query information is displayed on the dialogue interface, wherein the reply information is obtained by inferring the sub-query information corresponding to the verification result when the verification result corresponding to the sub-query information is greater than the verification result threshold, and the verification result is obtained by verifying the sub-reply information matching the sub-query information, and is used to characterize the feasibility of the sub-query information for the inferred query information, thereby achieving a technical effect of improving the efficiency of information reasoning and solving the technical problem of low efficiency of information reasoning.

[0143] The above method of this embodiment is further introduced below.

[0144] As an optional embodiment, the query information is multimodal information, and the type of multimodal information includes at least one of the following: text information containing character information, video frame information containing frame image information, and audio information; the type of reply information includes at least one of the following: text information, image information, video information, and voice information.

[0145] In this embodiment, the query information may be multimodal information. The multimodal information may include at least one of the following: text information including character information, video frame information including frame image information, and audio information. The reply information may include at least one of the following: text information, image information, video information, and voice information.

[0146] It should be noted that the types of query information sent by the above-mentioned users and the types of reply information replied by the large model are only examples and are not specifically limited here.

[0147] In an embodiment of the present disclosure, after detecting the presence of query information in the interactive interface of a certain client, the query information can be transmitted from the client to the server, where the query information is decomposed and processed to determine sub-query information that has a logical association with each other in different reasoning stages. And each sub-query information can be replied to to obtain corresponding sub-reply information. In order to ensure the feasibility of the determined reply information, each sub-reply information can be verified, and each sub-reply information corresponds to a verification result. Whether the query information is feasible is determined by judging whether the verification result is greater than the verification result threshold. If the verification result is greater than the verification result threshold, it means that the query information is highly feasible, and the sub-query information can be reasoned to determine the reply information of the query information. Taking into account the low efficiency of reasoning logical problems with logical relationships in related technologies due to the lack of cognitive ability of large models, the embodiments of the present disclosure can divide the original query information into multiple sub-query information with logical associations according to the before and after reasoning stages. By responding to the sub-query information at different reasoning stages, the reply information of the query information can be finally determined, thereby greatly improving the accuracy of the large model in solving logical reasoning and other problems, thereby achieving the technical effect of improving the efficiency of information reasoning and solving the technical problem of low efficiency of information reasoning.

[0148] The embodiment of the present disclosure also provides a method for generating a model on the model training side. FIG4 is a flow chart of a method for generating a model according to an embodiment of the present disclosure. As shown in FIG4 , the method may include the following steps:

[0149] Step S402: Obtain query information sample.

[0150] In the technical solution provided in the above step S402 of the present disclosure, query information samples for training the first large model can be obtained.

[0151] In step S404, the first large model is trained using the sub-query information samples corresponding to the query information samples, wherein the first large model is used to decompose the input query information to obtain sub-query information at different reasoning stages, and the sub-query information at different reasoning stages has a logical association relationship.

[0152] Alternatively, a query information sample can be obtained and input into a generative model for analysis to obtain corresponding sub-query information samples, and the loss degree of the sub-query information sample can be determined. The generative model can then be trained using the loss degree and the sample set to obtain a first large model, where the sub-query information sample can be used to represent the text generated by decomposition based on the query information sample. The sample set can be a context example.

[0153] In the technical solution provided in the above step S402 of the present disclosure, after obtaining the query information sample, the sub-query information sample corresponding to the query information sample can be used to train a first large model, wherein the first large model can be used to decompose and process the input query information to obtain sub-query information of different reasoning stages, and there is a logical association between the sub-query information that does not pass the reasoning stage.

[0154] Step S406: Obtain a response information sample corresponding to the query information sample.

[0155] In the technical solution provided in the above step S406 of the present disclosure, a response information sample corresponding to the query information sample may also be obtained.

[0156] Step S408: Acquire a verification result sample corresponding to the reply information sample, wherein the verification result sample is used to at least represent a classification result of the reply information sample.

[0157] In the technical solution provided in the above step S408 of the present disclosure, after obtaining the reply information sample corresponding to the query information sample, the sample result sample corresponding to the reply information sample can be obtained, wherein the verification result sample can be used to at least represent the classification result of the reply information sample.

[0158] Optionally, query information samples and corresponding response information samples are obtained, and the query information samples and response information samples can be input into the generative model for analysis to obtain corresponding verification result samples. The loss degree of the verification result samples can be determined, and the verification result samples can be used to train the generative model to obtain the second largest model.

[0159] Step S410: Use the verification result sample to train to obtain the second largest model, wherein the second largest model is used to verify the sub-response information that matches the sub-query information, and obtain the verification result corresponding to the sub-query information. The verification result is used to characterize the feasibility of the sub-query information for inferring the query information, and the sub-query information corresponding to the verification result greater than the verification result threshold is used to infer the response information corresponding to the query information.

[0160] In the technical solution provided in step S410 above, after obtaining a sample result sample corresponding to the response information sample, a second large model can be trained using the verification result sample. The second large model can be used to verify the sub-response information that matches the sub-query information, obtaining a verification result corresponding to the sub-query information. The verification result can be used to indicate the feasibility of the sub-query information for inferring the query information. Sub-query information corresponding to a verification result greater than a verification result threshold is used to infer the response information corresponding to the query information.

[0161] Optionally, after verifying the sub-query information and sub-reply information and obtaining a verification result, the magnitude relationship between the verification result and the verification result threshold can be determined. If the verification result is less than or equal to the verification result threshold, it can be indicated that the correctness of the obtained sub-query information and sub-reply information is low. If the reply information of the final query information is inferred based on the sub-reply information at this time, the accuracy cannot be guaranteed. If the verification result is greater than the verification result threshold, it can be indicated that the correctness of the obtained sub-query information and sub-reply information is high. In this case, the reply information of the final query information can be inferred based on the sub-reply information to ensure the accuracy of the reply information.

[0162] The above method of this embodiment is further introduced below.

[0163] As an optional implementation, sub-query information samples corresponding to query information samples are used to train a first large model, including: determining the loss degree of the sub-query information samples; using the loss degree and the sample set to train a generative model to obtain the first large model, wherein the sample set includes sub-query information corresponding to different query information, and wherein the generative model is a pre-trained generative large model.

[0164] In this embodiment, when training a first large model using sub-query information samples corresponding to a query information sample, the loss degree of the sub-query information sample can be determined. The loss degree and the sample set are then used to train a generative model to obtain the first large model. The sample set can include sub-query information corresponding to different query information. The generative model is a pre-trained generative large model.

[0165] Alternatively, SFT has been demonstrated to be effective in understanding user intent. Therefore, in the disclosed embodiment, the intuitive system can utilize contextual examples to decompose query Q into sub-questions. That is, the intuitive system can decompose query information samples of complex questions into corresponding sub-query information samples. Because a generative model can be used as the intuitive system, during training of the intuitive system, loss calculations can be performed on the sub-query information samples generated based on the query information samples to determine the loss degree of the sub-query information. The loss degree and sample set can then be used to train the intuitive system.

[0166] Optionally, during the autoregressive calculation, the generated text without the given context can be used, that is, the loss is calculated only for the subquery information sample. For example, given a sample of length N, it can be represented by X, where X = {x1, ...x i ,…x n}. The sequence length of the context example can be defined as M. The standard language modeling objective is used to maximize the following likelihood function: This trains the intuitive system.

[0167] As an optional implementation, the second largest model is obtained by training the verification result samples, including: determining the loss degree of the verification result samples; using the loss degree of the verification result samples to train the generative model to obtain the second largest model, wherein the generative model is a pre-trained generative large model.

[0168] In this embodiment, in the process of training the second largest model using the verification result samples, the loss degree of the verification result samples can be determined, and the loss degree of the verification result samples can be used to connect the generative model to obtain the second largest model, wherein the generative model is a pre-trained generative large model.

[0169] Optionally, in the process of training the reflective system, the same training method as the intuitive system can be adopted, using positive and negative samples to allow the generative model to generate classification results from the positive and negative samples, that is, to generate corresponding verification result samples, by determining the loss degree of the verification result samples.

[0170] Alternatively, since the reflection system mainly focuses on the judgment of state s, the following loss function can be used to determine the loss of the verification result sample:

[0171] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0172] It should be noted that for the aforementioned method embodiments, for simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present disclosure is not limited by the order of the actions described, because according to the present disclosure, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present disclosure.

[0173] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present disclosure.

[0174] Example 2

[0175] According to an embodiment of the present disclosure, an information reasoning system is also provided. FIG5 is a schematic diagram of an information reasoning system according to an embodiment of the present disclosure. As shown in FIG5 , the information reasoning system may include: an information generation terminal 502 and an information verification terminal 504 .

[0176] The information generation terminal 502 is used to detect the query information to be inferred, decompose the query information, obtain sub-query information of different reasoning stages, and determine the sub-response information that matches the sub-query information of different reasoning stages, wherein the sub-query information of different reasoning stages has a logical association relationship.

[0177] In the technical solution provided by the above-mentioned information generation terminal 502 of the present disclosure, the query information to be inferred can be detected by the information generation terminal 502, and the query information can be decomposed and processed to obtain sub-query information of different reasoning stages, and the sub-response information corresponding to each sub-query information can be determined respectively, wherein the sub-query information of different reasoning stages has a logical association relationship.

[0178] Optionally, the information generating terminal 502 performs reasoning and analysis on the query information input on the interactive interface that can describe the query question.

[0179] Optionally, the information generating terminal 502 may divide the query information into a plurality of sub-query information according to the logical association relationship in the reasoning stage.

[0180] Optionally, the information generation terminal 502 can perform reasoning and analysis on each of the sub-query information contained in each reasoning stage obtained by decomposing the query information, thereby determining the solution process and final result for each sub-query information. In other words, the sub-answer information corresponding to each sub-query information can be determined through reasoning and analysis. The sub-answer information can then be transmitted to the information verification terminal 504.

[0181] The information verification terminal 504 is used to verify the sub-response information matched by the sub-query information, obtain the verification result corresponding to the sub-query information, and in response to the verification result being greater than the verification result threshold, infer the sub-query information corresponding to the verification result to obtain the reply information corresponding to the query information, wherein the verification result is used to characterize the feasibility of the sub-query information for the inferred query information.

[0182] In the technical solution provided by the information verification terminal 504 of the present disclosure, the sub-query information can be verified by the information verification terminal 504 to determine a verification result. When the verification result is greater than a verification result threshold, the sub-query information corresponding to the verification result can be inferred to obtain response information corresponding to the query information. The verification result can be used to indicate the feasibility of the sub-query information for the inferred query information.

[0183] Optionally, in order to ensure the accuracy of the reply information of the query information finally determined, it is necessary to evaluate the sub-reply information obtained by solving each sub-query information obtained by decomposition through the information verification terminal 504 to evaluate the acceptability of the sub-reply information. This method can be used to ensure the accuracy of the determined sub-reply information, thereby ensuring the accuracy of the reply information of the final query information.

[0184] Optionally, after verifying the sub-query information and sub-reply information and obtaining a verification result, the magnitude relationship between the verification result and the verification result threshold can be determined. If the verification result is less than or equal to the verification result threshold, it can be indicated that the correctness of the obtained sub-query information and sub-reply information is low. If the reply information of the final query information is inferred based on the sub-reply information at this time, the accuracy cannot be guaranteed. If the verification result is greater than the verification result threshold, it can be indicated that the correctness of the obtained sub-query information and sub-reply information is high. In this case, the reply information of the final query information can be inferred based on the sub-reply information to ensure the accuracy of the reply information.

[0185] The information reasoning system disclosed herein detects query information to be reasoned through an information generation end, decomposes the query information to obtain sub-query information at different reasoning stages, and determines sub-answer information that matches the sub-query information at different reasoning stages, wherein the sub-query information at different reasoning stages has a logical association relationship. An information verification end verifies the sub-answer information that matches the sub-query information, obtains a verification result corresponding to the sub-query information, and, in response to the verification result being greater than a verification result threshold, performs reasoning on the sub-query information corresponding to the verification result to obtain answer information corresponding to the query information, wherein the verification result indicates the feasibility of the sub-query information for reasoning the query information. This achieves the technical effect of improving the efficiency of information reasoning and solves the technical problem of low information reasoning efficiency.

[0186] Example 3

[0187] Currently, large language models have achieved impressive results in a variety of tasks, including idea generation, dialogue systems, and brainstorming. However, large models have certain limitations in complex tasks such as logical reasoning and solving difficult mathematical problems, especially those involving abstract concepts or multi-step reasoning.

[0188] With the continuous advancement of deep learning in tasks such as natural language processing and machine translation, there is growing interest in applying deep learning to natural language processing. This has led to the emergence of large language models (such as GPT-3.5), which have achieved significant breakthroughs in tasks such as text generation, sentiment analysis, and dialogue systems. Large language models are typically pre-trained on large amounts of text data and then fine-tuned to optimize for specific tasks to produce high-quality text output. These models have strong natural language understanding and generation capabilities, capable of understanding context and generating coherent language, and have therefore attracted considerable attention in the field of natural language processing. Efforts are underway to improve the performance of large language models by continuously refining model structures and training methods to better meet the growing demands of applications. Developments in this field present new opportunities and challenges for natural language processing and artificial intelligence research.

[0189] However, in related technologies, it is still difficult for language models to solve complex logical reasoning problems and mathematical problems. In addition, traditional language models lack cognitive capabilities. When dealing with problems involving lengthy reasoning chains or multi-step solutions, it is important to evaluate the question and its current answer. However, the deployment and reasoning costs of large language models are relatively high, especially when using reasoning enhancement techniques without parameter updates. These techniques require a large amount of context and multi-step answer generation, further increasing the reasoning cost and time. Therefore, there is still a technical problem of low efficiency of information reasoning.

[0190] Optionally, the present disclosure provides a complex task reasoning method for large language models, which solves the technical problem of low efficiency of information reasoning. It is different from the traditional solution in which poor cognitive ability leads to long reasoning time for complex tasks and low efficiency of information reasoning. It solves the technical problem of low efficiency of information reasoning.

[0191] In an embodiment of the present disclosure, after detecting the presence of query information in the interactive interface of a certain client, the query information can be transmitted from the client to the server, where the query information is decomposed and processed to determine sub-query information that has a logical association with each other in different reasoning stages. And each sub-query information can be replied to to obtain corresponding sub-reply information. In order to ensure the feasibility of the determined reply information, each sub-reply information can be verified, and each sub-reply information corresponds to a verification result. Whether the query information is feasible is determined by judging whether the verification result is greater than the verification result threshold. If the verification result is greater than the verification result threshold, it means that the query information is highly feasible, and the sub-query information can be reasoned to determine the reply information of the query information.

[0192] Since the embodiments of the present disclosure take into account the low efficiency of reasoning logical problems with logical relationships in related technologies due to the lack of cognitive ability of large models, the original query information can be divided into multiple sub-query information with logical associations according to the before and after reasoning stages. By responding to the sub-query information at different reasoning stages, the reply information of the query information can be finally determined, thereby greatly improving the accuracy of the large model in solving logical reasoning and other problems, and further achieving the technical effect of improving the efficiency of information reasoning, and solving the technical problem of low efficiency of information reasoning.

[0193] The above method of this embodiment is further introduced below.

[0194] In this embodiment, for complex task reasoning of lightweight large models, a smaller-scale model (7B) is used to construct a dual-system generation reasoning tree, which effectively enhances the model's ability to answer complex mathematical problems and logical reasoning problems. A method for solving complex mathematical problems with large models is proposed. This method is based on human cognitive theory and imitates the process of human cognition through two systems: the intuitive system and the reflective system. The intuitive system is responsible for generating multiple decomposition hypotheses of the original problem, and the reflective system verifies the hypotheses generated by the intuitive system and selects more likely hypotheses for subsequent generation until the final result is reached. Through the iterative generation of the above-mentioned dual system, the problem-solving accuracy of the large model can be improved.

[0195] In related technologies, models like ChatGPT or GPT-4, which are used to solve logical reasoning or mathematical problems, require contextual examples. However, these models have high inference costs (time and space). For example, the model has 175 bytes of parameters and is not open source. Smaller models, such as LLaMA-7B, do not achieve the expected accuracy for complex tasks.

[0196] Optionally, in order to lightweight complex task reasoning, a smaller-scale model can be used to build a dual-system generative reasoning tree to greatly enhance the large model's ability to answer complex mathematical and logical reasoning problems.

[0197] Optionally, the cognitive tree uses a smaller and more open-source model to solve the problem of high model inference cost, and the inference tree is a process of generating an inference tree using a dual system to enhance the problem of insufficient inference ability of the large model itself.

[0198] Optionally, the reasoning framework can adopt an iterative method to build a cognitive tree structure based on human thinking patterns. The root node of the cognitive tree can represent the initial query, that is, the query information, and the leaf nodes can contain simple questions that are directly answered, that is, the sub-query information.

[0199] In this embodiment, FIG6 is a schematic diagram of a cognitive tree according to an embodiment of the present disclosure. As shown in FIG6 , when faced with complex task reasoning, such as logical reasoning problems and mathematical problems, the client can issue query information that requires a reply from the large model based on the above problems. The query information can be transmitted to the large model of the dual system. Drawing on the human thinking mode, a cognitive tree structure as shown in FIG6 can be proposed. The large model can be decomposed to construct two systems: an intuitive system 601 and a reflective system 602. The intuitive system 601 is used to generate a hypothesis for the decomposition of the original problem, that is, the query information can be decomposed into sub-query information using the intuitive system. As shown in FIG6 , the query information can be decomposed into sub-query information A and sub-query information B. Sub-query information A cannot be divided any further, and sub-query information B can be further divided into sub-query information B1 and sub-query information B2. The reflective system can be used to verify the correctness of the hypothesis to guide the subsequent generation of the intuitive system. That is, the reflective system 602 can be used to evaluate the sub-query information decomposed by the intuitive system. Only after the evaluation meets the requirements can subsequent operations be performed through the intuitive system. That is, the intuitive system is subsequently used to respond to each sub-query information, obtain corresponding sub-response information, and finally obtain response information for the query information to answer more complex logical reasoning and mathematical problems for the client.

[0200] In this disclosed embodiment, a dual-system iteratively generates an inference tree to enhance the reasoning capabilities of large models. The innovation of this method is that it designs a new inference framework for large language models, enhancing the reasoning capabilities of large models on complex mathematical and logical reasoning problems.

[0201] Alternatively, the generative capabilities of the intuitive system are fundamental to building a cognitive tree. Therefore, a decoder-only model is selected as the intuitive system. The capabilities of the intuitive system are enhanced through contextual methods. Query Q is defined as the ultimate goal of a logical reasoning problem or a mathematical problem.

[0202] Alternatively, in the case of a logical reasoning problem, decomposition D involves further decomposing the query information into smaller problems, and the final goal can be achieved by reasoning on the above decompositions.

[0203] Alternatively, in the case of mathematical problems, decomposition D refers to a subproblem derived from the original problem, the solution of which helps solve the entire original problem.

[0204] Optionally, the decomposition set can represent the decomposition set of examples in the training set. From the inference decomposition set, k examples can be retrieved (e.g., query: Q; decomposition: query D), and then the above examples can be used as context for the model input. The output can be generated as y~f θ (y|x,z 1…k), where [y]~f θ (y|x,z 1…k ) can be used to represent a continuous language sequence. z can be used to represent the k examples retrieved from the decomposition set Z, Z = {z1,…z L}.

[0205] Optionally, after obtaining the query information sent by the current client, the query information in the sample set can be traversed and the similarity between each query information and the query information sent by the current client can be determined. If the similarity reaches a similarity threshold, it can be considered that the two are highly similar. In this case, the query information in the sample set that is highly similar to the query information sent by the current client, as well as the sub-query information corresponding to the query information, can be used as the corresponding sub-query information of the query information sent by the current client in the inference phase.

[0206] Alternatively, an intuitive system can be used to obtain a representation of the current query and calculate its cosine similarity with the representations of other queries in the collection. Specifically, the identifiers of the queries included in the sample set can be traversed. During this traversal, the cosine similarity between each identifier in the sample set and the identifier of the query sent by the client can be determined. K queries that meet the requirements can then be retrieved from the collection.

[0207] Optionally, during the training of an intuitive system, SFT has been demonstrated to be effective in recognizing user intent. Therefore, in the disclosed embodiment, the intuitive system can utilize contextual examples to decompose query Q into sub-questions. That is, the intuitive system can decompose query information samples of complex questions into corresponding sub-query information samples. Because a generative model can be used as the intuitive system, during the training of the intuitive system, loss calculations can be performed on the sub-query information samples generated based on the query information samples to determine the loss degree of the sub-query information. The loss degree and sample set can then be used to train the intuitive system.

[0208] Optionally, during the autoregressive calculation, the generated text without the given context can be used, that is, the loss is calculated only for the subquery information sample. For example, given a sample of length N, it can be represented by X, where X = {x1, ...x i ,…x n}. The sequence length of the context example can be defined as M. The standard language modeling objective is used to maximize the following likelihood function: This trains the intuitive system.

[0209] Alternatively, the reflective system functions differently from the intuitive system. While the intuitive system relies on quick intuition for generation, the reflective system's role is to evaluate the intuitive system's output to determine its acceptability. The reflective system can verify its output in two ways: by verifying intermediate processes and by verifying the overall chain of reasoning.

[0210] Alternatively, for verification of the intermediate process, given the current state S (query: Q and decomposition: D), a reflective system with the same model architecture as the intuitive system can be used to generate a score v that verifies the current state, where v can be expressed as V(f θ ,s)~f θ (v|s).

[0211] Alternatively, for verification of the entire inference chain, the completed inference chain can be given as S = {s1,…,s i ,…,s n}, we can use the reflection system to produce an overall fraction o, which can be expressed as O(f θ ,S)~f θ (o|S).

[0212] In the disclosed embodiment, the reflective system is different from the intuitive system in that its important task is to evaluate and verify the feasibility of the entire reasoning chain of the current state, rather than generating quick hypotheses like the intuitive system. This evaluation process helps to determine whether the generated hypotheses and reasoning process are reasonable, thereby achieving the technical effect of improving the accuracy of information reasoning of large models.

[0213] Optionally, in the process of training the reflective system, the same training method as the intuitive system can be used, using positive and negative samples to allow the model to generate classification results. Since the reflective system mainly focuses on the judgment of state s, the loss function can be defined as follows:

[0214] For example, Figure 7 is a schematic diagram of a complex task reasoning process according to an embodiment of the present disclosure. As shown in Figure 7, if the query information that the client wants to query the large model is "Wen earns 12 yuan per hour as a nanny. Yesterday, she only worked as a nanny for 50 minutes. How much did she earn?", the client's query information can be sent to the intuitive system in the server. The intuitive system can decompose the query information according to the reasoning stage. If the reasoning stage includes reasoning stage 1 and reasoning stage 2, the intuitive system can decompose the query information in reasoning stage 1 into three sub-queries and corresponding sub-answer information, 1a, 1b, and 1c. Among them, sub-query information and sub-answer information 1a can be "How much does Wen earn per minute? Wen earns 12 / 60 = 0.2 yuan per minute"; sub-query information and sub-answer information 1b can be "How much does the nanny charge per hour? The nanny charges 12 yuan per hour"; sub-query information and sub-answer information 1c can be "How many hours are equal to 50 minutes? 50 minutes equals 1 hour." Through the intuitive system, the reasoning stage 2 can be decomposed into three sub-query information and corresponding sub-answer information, 2a, 2b and 2c, among which the sub-query information and sub-answer information 2a can be "How long did Wen work yesterday? Wen worked 1 hour yesterday"; the sub-query information and sub-answer information 2b can be "How much money did Wen earn? She worked for 50 minutes and earned 0.2x50=10 yuan"; the sub-query information and sub-answer information 2c can be "How many minutes did Wen work as a nanny yesterday? 50 minutes".

[0215] For another example, as shown in Figure 7, after the intuitive system decomposes the sub-query information and determines the sub-answer information, the sub-query information and sub-answer information can be transmitted to the reflective system. The reflective system evaluates the sub-answer information and sub-query information determined by the intuitive system. For example, for 1a, the reflective system evaluates it as "certain"; for 1b, it evaluates it as "possible"; and for 1c, it evaluates it as "impossible." For 2a, the reflective system evaluates it as "impossible"; for 2b, it evaluates it as "certain"; and for 2c, it evaluates it as "possible." The reflective system can determine the current state score and the overall score. Specifically, the reflective system can evaluate each sub-query information and sub-answer information at each reasoning stage, as well as the entire reasoning chain of the intuitive system. Therefore, if the evaluation deems the current sub-answer information and sub-query information poor, it can be returned to the intuitive system for re-decomposition. The resulting answer to the query is "Wen earned 10 yuan yesterday." This answer can be transmitted via the network to the corresponding client device's interactive interface for display.

[0216] It should be noted that the preferred implementation scheme involved in the above embodiments of the present disclosure is the same as the solution provided in Example 1, as well as the application scenario and implementation process, but is not limited to the solution provided in Example 1.

[0217] Example 4

[0218] According to an embodiment of the present disclosure, an information reasoning device for implementing the information reasoning method shown in FIG. 2 is also provided.

[0219] FIG8 is a schematic diagram of an information inference device according to an embodiment of the present disclosure. As shown in FIG8 , the information inference device 800 may include: a detection unit 802 , a processing unit 804 , a determination unit 806 , a verification unit 808 , and an inference unit 810 .

[0220] The detection unit 802 is used to detect the query information to be inferred.

[0221] The processing unit 804 is configured to decompose the query information to obtain sub-query information at different reasoning stages, wherein the sub-query information at different reasoning stages has a logical association relationship.

[0222] The determining unit 806 is configured to determine sub-answer information that matches the sub-query information at different reasoning stages.

[0223] The verification unit 808 is used to verify the sub-response information matched by the sub-query information, and obtain a verification result corresponding to the sub-query information, wherein the verification result is used to represent the feasibility of the sub-query information to the inference query information.

[0224] The reasoning unit 810 is configured to, in response to the verification result being greater than the verification result threshold, infer the sub-query information corresponding to the verification result to obtain reply information corresponding to the query information.

[0225] It should be noted that the detection unit 802, processing unit 804, determination unit 806, verification unit 808, and inference unit 810 described above correspond to steps S202 to S210 in Example 1. The examples and application scenarios implemented by the five units and the corresponding steps are the same, but are not limited to the contents disclosed in Example 1. It should be noted that the above-mentioned units can be hardware components or software components stored in a memory and processed by one or more processors. The above-mentioned units can also be part of the device and can be run in the computer terminal provided in Example 5.

[0226] According to an embodiment of the present disclosure, an information reasoning device for implementing the information reasoning method shown in FIG. 3 is also provided.

[0227] FIG9 is a schematic diagram of another information inference device according to an embodiment of the present disclosure. As shown in FIG9 , the information inference device 900 may include: a first display unit 902 , a second display unit 904 , and a third display unit 906 .

[0228] The first display unit 902 is configured to display sub-query information of the query information at different reasoning stages on the dialogue interface in response to the query information to be reasoned received in the dialogue interface, wherein the sub-query information at different reasoning stages has a logical association relationship.

[0229] The second display unit 904 is configured to display, on the dialogue interface, sub-answer information that matches the sub-query information at different reasoning stages.

[0230] The third display unit 906 is used to display the reply information corresponding to the query information on the dialogue interface, wherein the reply information is obtained by inferring the sub-query information corresponding to the verification result when the verification result corresponding to the sub-query information is greater than the verification result threshold, and the verification result is obtained by verifying the sub-reply information matching the sub-query information, and is used to represent the feasibility of the sub-query information to the inferred query information.

[0231] It should be noted that the first display unit 902, the second display unit 904, and the third display unit 906 correspond to steps S302 to S306 in Example 1. The three units and the corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in Example 1. It should be noted that the above units can be hardware components or software components stored in a memory and processed by one or more processors. The above units can also be part of the device and can be run in the computer terminal provided in Example 5.

[0232] According to an embodiment of the present disclosure, a model generation device for implementing the model generation method shown in FIG. 4 is also provided.

[0233] FIG10 is a schematic diagram of a model generation device according to an embodiment of the present disclosure. As shown in FIG10 , the model generation device 1000 may include: a first acquisition unit 1002 , a first training unit 1004 , a second acquisition unit 1006 , a third acquisition unit 1008 , and a second training unit 1010 .

[0234] The first acquiring unit 1002 is configured to acquire a query information sample.

[0235] The first training unit 1004 is used to train a first large model using sub-query information samples corresponding to the query information samples, wherein the first large model is used to decompose the input query information to obtain sub-query information at different reasoning stages, and the sub-query information at different reasoning stages has a logical association relationship.

[0236] The second acquiring unit 1006 is configured to acquire a response information sample corresponding to the query information sample.

[0237] The third obtaining unit 1008 is configured to obtain a verification result sample corresponding to the reply information sample, wherein the verification result sample is used to at least represent a classification result of the reply information sample.

[0238] The second training unit 1010 is used to train a second large model using the verification result sample, wherein the second large model is used to verify the sub-response information matched by the sub-query information to obtain a verification result corresponding to the sub-query information, and the verification result is used to characterize the feasibility of the sub-query information for inferring the query information, and the sub-query information corresponding to the verification result greater than the verification result threshold is used to infer the response information corresponding to the query information.

[0239] It should be noted that the first acquisition unit 1002, the first training unit 1004, the second acquisition unit 1006, the third acquisition unit 1008, and the second training unit 1010 described above correspond to steps S402 to S410 in Example 1. The examples and application scenarios implemented by the five units and the corresponding steps are the same, but are not limited to the contents disclosed in Example 1. It should be noted that the above units can be hardware components or software components stored in a memory and processed by one or more processors. The above units can also be run as part of the device in the computer terminal provided in Example 5.

[0240] In the above-described device, when a client requests a large model to respond to a query, it can enter query information describing the query content on the client's interactive interface. After detecting the presence of query information in a client's interactive interface, the query information can be transmitted from the client to the server, where the query information is decomposed and processed to determine sub-query information with logical relationships between them at different reasoning stages. Each sub-query information can then be responded to, resulting in a corresponding sub-response information. To ensure the feasibility of the determined response information, each sub-response information can be verified, and each sub-response information can be verified to determine the feasibility of the query information. The feasibility of the query information is determined by determining whether the verification result is greater than a verification result threshold. If the verification result is greater than the verification result threshold, it indicates that the query information is highly feasible, and reasoning can be performed on the sub-query information to determine the response information for the query information. Taking into account the low efficiency of reasoning logical problems with logical relationships in related technologies due to the lack of cognitive ability of large models, the embodiments of the present disclosure can divide the original query information into multiple sub-query information with logical associations according to the before and after reasoning stages. By responding to the sub-query information at different reasoning stages, the reply information of the query information can be finally determined, thereby greatly improving the accuracy of the large model in solving logical reasoning and other problems, thereby achieving the technical effect of improving the efficiency of information reasoning and solving the technical problem of low efficiency of information reasoning.

[0241] Example 5

[0242] The embodiment of the present disclosure may provide a computer terminal, which may be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal may also be replaced by a terminal device such as a mobile terminal.

[0243] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of a computer network.

[0244] In this embodiment, the above-mentioned computer terminal can execute the program code of the following steps in the information reasoning method: detecting the query information to be inferred; decomposing the query information to obtain sub-query information of different reasoning stages, wherein the sub-query information of different reasoning stages have a logical association relationship; determining the sub-reply information that matches the sub-query information of different reasoning stages respectively; verifying the sub-reply information that matches the sub-query information to obtain a verification result corresponding to the sub-query information, wherein the verification result is used to characterize the feasibility of the sub-query information for reasoning the query information; in response to the verification result being greater than the verification result threshold, inferring the sub-query information corresponding to the verification result to obtain the reply information corresponding to the query information.

[0245] Optionally, Figure 11 is a block diagram of a computer terminal according to an embodiment of the present disclosure. As shown in Figure 11, the computer terminal A may include: one or more (only one is shown in the figure) processors 1102, a memory 1104, a storage controller, and a peripheral interface, wherein the peripheral interface is connected to a radio frequency module, an audio module, and a display.

[0246] Among them, the memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the information reasoning method and device in the embodiments of the present disclosure. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned information reasoning method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely located relative to the processor, and these remote memories may be connected to the computer terminal A via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0247] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: based on the query information, search for sub-query information of different reasoning stages in the sample set, wherein the sample set includes sub-query information corresponding to different query information.

[0248] Optionally, the processor may further execute program code of the following steps: determining identification information of the query information; and determining sub-query information whose identification information matches the identification information of the query information in a sample set corresponding to the inference stage as sub-query information of the inference stage.

[0249] Optionally, the processor may further execute program code of the following steps: retrieving sub-query information whose identification information has a similarity greater than a similarity threshold to the identification information of the query information in the sample set corresponding to the inference stage; and determining the retrieved sub-query information as the sub-query information of the inference stage.

[0250] Optionally, the processor may further execute program code of the following steps: using the first large model to analyze the query information to obtain identification information of the query information.

[0251] Optionally, the processor may also execute the program code of the following steps: using the first large model to analyze the sample set corresponding to the inference stage and the identification information of the query information to obtain sub-query information whose identification information matches the identification information of the query information; and using the first large model to determine the matched sub-query information as the sub-query information of the inference stage.

[0252] Optionally, the processor may also execute the program code of the following steps: obtaining a query information sample; inputting the query information sample into the generation model for analysis to obtain a corresponding sub-query information sample; determining the loss degree of the sub-query information sample; and using the loss degree and the sample set to train the generation model to obtain a first large model.

[0253] Optionally, the processor may further execute program code of the following steps: using the first large model to analyze sub-query information at different reasoning stages, and obtaining sub-answer information matching the sub-query information at different reasoning stages.

[0254] Optionally, the above-mentioned processor can also execute the program code of the following steps: when running to the current reasoning stage in different reasoning stages, verify the sub-response information that matches the sub-query information of the current reasoning stage to obtain the sub-verification result of the current reasoning stage, wherein the verification result includes the sub-verification result; or, verify the sub-response information of different reasoning stages to obtain the overall verification result of different reasoning stages, wherein the verification result includes the overall verification result.

[0255] Optionally, the above-mentioned processor can also execute the program code of the following steps: use the second largest model to verify the sub-response information of the current reasoning stage to obtain the sub-verification result; use the second largest model to verify the sub-response information of different reasoning stages to obtain the overall verification result.

[0256] Optionally, the above-mentioned processor can also execute the program code of the following steps: obtain query information samples and corresponding reply information samples; input the query information samples and reply information samples into the generation model for analysis to obtain corresponding verification result samples; determine the loss degree of the verification result samples; use the verification result samples to train the generation model to obtain the second largest model.

[0257] Optionally, the above-mentioned processor can also execute the program code of the following steps: in the sample set corresponding to the current reasoning stage, searching for sub-query information of the current reasoning stage based on the query information; in the case where different reasoning stages include the next reasoning stage of the current reasoning stage, in the sample set corresponding to the next reasoning stage, searching for sub-query information of the next reasoning stage based on the sub-verification result and query information of the current reasoning stage.

[0258] Optionally, the above-mentioned processor can also execute the program code of the following steps: determine the query information as the root node, determine the sub-query information as the leaf node, and establish a target tree structure; infer the sub-query information corresponding to the verification result to obtain the reply information corresponding to the query information, including: in the target tree structure, infer the sub-query information corresponding to the verification result to obtain the reply information corresponding to the query information.

[0259] As another optional example, the processor can call the information and application stored in the memory through the transmission device to perform the following steps: in response to the query information to be inferred received in the dialogue interface, display the sub-query information of the query information at different reasoning stages on the dialogue interface, wherein the sub-query information at different reasoning stages has a logical association relationship; display the sub-reply information that matches the sub-query information at different reasoning stages on the dialogue interface; display the reply information corresponding to the query information on the dialogue interface, wherein the reply information is obtained by inferring the sub-query information corresponding to the verification result when the verification result corresponding to the sub-query information is greater than the verification result threshold, and the verification result is obtained by verifying the sub-reply information matching the sub-query information, and is used to characterize the feasibility of the sub-query information for reasoning the query information.

[0260] Optionally, the above-mentioned processor can also execute the program code of the following steps: the query information is multimodal information, the type of the multimodal information includes at least one of the following: text information containing character information, video frame information containing frame image information, audio information, and the type of the reply information includes at least one of the following: text information, image information, video information and voice information.

[0261] As another optional example, the processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain a query information sample; use the sub-query information sample corresponding to the query information sample to train a first large model, wherein the first large model is used to decompose the input query information to obtain sub-query information at different reasoning stages, and the sub-query information at different reasoning stages has a logical association relationship; obtain a reply information sample corresponding to the query information sample; obtain a verification result sample corresponding to the reply information sample, wherein the verification result sample is used to at least represent the classification result of the reply information sample; use the verification result sample to train a second large model, wherein the second large model is used to verify the sub-reply information matched by the sub-query information to obtain a verification result corresponding to the sub-query information, and the verification result is used to characterize the feasibility of the sub-query information for reasoning the query information, and the sub-query information corresponding to the verification result greater than the verification result threshold is used to infer the reply information corresponding to the query information.

[0262] Optionally, the processor may also execute the program code of the following steps: determining the loss degree of the sub-query information sample; training a generative model using the loss degree and the sample set to obtain a first large model, wherein the sample set includes sub-query information corresponding to different query information, and wherein the generative model is a pre-trained generative large model.

[0263] Optionally, the processor may also execute the program code of the following steps: determining the loss degree of the verification result sample; using the loss degree of the verification result sample to train a generative model to obtain a second large model, wherein the generative model is a pre-trained generative large model.

[0264] According to the embodiment of the present disclosure, when the client has the need to request a large model to respond to the content to be queried, query information describing the content to be queried can be input on the interactive interface of the client. After detecting the existence of query information in the interactive interface of a certain client, the query information can be transmitted from the client to the server, and the query information can be decomposed and processed in the server to determine the sub-query information that has a logical relationship with each other in different reasoning stages. And each sub-query information can be responded to to obtain the corresponding sub-response information. In order to ensure the feasibility of the determined reply information, each sub-response information can be verified, and each sub-response information corresponds to a verification result. Whether the query information is feasible is determined by judging whether the verification result is greater than the verification result threshold. If the verification result is greater than the verification result threshold, it means that the query information is highly feasible, and the sub-query information can be inferred to determine the reply information of the query information. Taking into account the low efficiency of reasoning logical problems with logical relationships in related technologies due to the lack of cognitive ability of large models, the embodiments of the present disclosure can divide the original query information into multiple sub-query information with logical associations according to the before and after reasoning stages. By responding to the sub-query information at different reasoning stages, the reply information of the query information can be finally determined, thereby greatly improving the accuracy of the large model in solving logical reasoning and other problems, thereby achieving the technical effect of improving the efficiency of information reasoning and solving the technical problem of low efficiency of information reasoning.

[0265] Those skilled in the art will appreciate that the structure shown in FIG11 is merely illustrative, and the computer terminal may also be a smartphone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile internet device (MID), a PAD, or other terminal device. FIG11 does not limit the structure of the aforementioned electronic devices. For example, the computer terminal A may include more or fewer components (such as a network interface, a display device, etc.) than those shown in FIG11, or may have a configuration different from that shown in FIG11.

[0266] A person skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0267] Example 6

[0268] The embodiment of the present disclosure further provides a computer-readable storage medium. Optionally, in this embodiment, the computer-readable storage medium can be used to store the program code executed by the information inference method provided in the first embodiment.

[0269] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0270] Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: detecting query information to be inferred; decomposing the query information to obtain sub-query information of different reasoning stages, wherein the sub-query information of different reasoning stages have a logical association relationship; determining sub-reply information that matches the sub-query information of different reasoning stages respectively; verifying the sub-reply information that matches the sub-query information to obtain a verification result corresponding to the sub-query information, wherein the verification result is used to characterize the feasibility of the sub-query information for reasoning the query information; in response to the verification result being greater than the verification result threshold, inferring the sub-query information corresponding to the verification result to obtain reply information corresponding to the query information.

[0271] Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: in response to query information to be inferred received in the dialogue interface, displaying sub-query information of the query information at different reasoning stages on the dialogue interface, wherein the sub-query information at different reasoning stages has a logical association relationship; displaying sub-reply information that matches the sub-query information at different reasoning stages on the dialogue interface; displaying reply information corresponding to the query information on the dialogue interface, wherein the reply information is obtained by inferring the sub-query information corresponding to the verification result when the verification result corresponding to the sub-query information is greater than the verification result threshold, and the verification result is obtained by verifying the sub-reply information matching the sub-query information, and is used to characterize the feasibility of the sub-query information for reasoning the query information.

[0272] Optionally, in this embodiment, the storage medium is configured to store program code for executing the following steps: obtaining a query information sample; using a sub-query information sample corresponding to the query information sample to train a first large model, wherein the first large model is used to decompose the input query information to obtain sub-query information at different reasoning stages, and the sub-query information at different reasoning stages has a logical association relationship; obtaining a reply information sample corresponding to the query information sample; obtaining a verification result sample corresponding to the reply information sample, wherein the verification result sample is used to at least represent a classification result of the reply information sample; using the verification result sample to train a second large model, wherein the second large model is used to verify the sub-reply information matched by the sub-query information to obtain a verification result corresponding to the sub-query information, and the verification result is used to characterize the feasibility of the sub-query information for reasoning the query information, and the sub-query information corresponding to the verification result greater than the verification result threshold is used to infer the reply information corresponding to the query information.

[0273] In an embodiment of the present disclosure, when a client has a need to request a large model to respond to the content to be queried, query information describing the content to be queried can be input on the interactive interface of the client. After detecting the presence of query information in the interactive interface of a certain client, the query information can be transmitted from the client to the server, where the query information is decomposed and processed to determine sub-query information that has a logical association with each other in different reasoning stages. And each sub-query information can be responded to to obtain corresponding sub-response information. In order to ensure the feasibility of the determined reply information, each sub-response information can be verified, and each sub-response information corresponds to a verification result. Whether the query information is feasible is determined by judging whether the verification result is greater than the verification result threshold. If the verification result is greater than the verification result threshold, it means that the query information is highly feasible, and the sub-query information can be inferred to determine the reply information of the query information. Taking into account the low efficiency of reasoning logical problems with logical relationships in related technologies due to the lack of cognitive ability of large models, the embodiments of the present disclosure can divide the original query information into multiple sub-query information with logical associations according to the before and after reasoning stages. By responding to the sub-query information at different reasoning stages, the reply information of the query information can be finally determined, thereby greatly improving the accuracy of the large model in solving logical reasoning and other problems, thereby achieving the technical effect of improving the efficiency of information reasoning and solving the technical problem of low efficiency of information reasoning.

[0274] Accordingly, an embodiment of the present disclosure further provides a computer program product. Optionally, in this embodiment, when the computer program product is executed by a processor, the method shown in the above method embodiment can be implemented.

[0275] Example 7

[0276] An embodiment of the present disclosure may provide an electronic device that may include a memory and a processor. Figure 12 is a block diagram of an electronic device according to an information reasoning method of an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0277] As shown in FIG12 , device 1200 includes a computing unit 1201, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 1202 or a computer program loaded from a storage unit 1208 into a random access memory (RAM) 1203. RAM 1203 may also store various programs and data required for the operation of device 1200. Computing unit 1201, ROM 1202, and RAM 1203 are interconnected via a bus 1204. An input / output (I / O) interface 1205 is also connected to bus 1204.

[0278] Various components in device 1200 are connected to I / O interface 1205, including: an input unit 1206, such as a keyboard, mouse, etc.; an output unit 1204, such as various types of displays, speakers, etc.; a storage unit 1208, such as a magnetic disk, optical disk, etc.; and a communication unit 1209, such as a network card, modem, wireless communication transceiver, etc. The communication unit 1209 allows device 1200 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0279] The computing unit 1201 can be a variety of general-purpose and / or specialized processing components with processing and computing capabilities. Some examples of the computing unit 1201 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 1201 performs the various methods and processes described above, such as the information reasoning method. For example, in some embodiments, the information reasoning method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as the storage unit 1208. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 1200 via the ROM 1202 and / or the communication unit 1209. When the computer program is loaded into RAM 1203 and executed by computing unit 1201, one or more steps of the information reasoning method described above may be performed. Alternatively, in other embodiments, computing unit 1201 may be configured to perform the information reasoning method in any other appropriate manner (e.g., by means of firmware).

[0280] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard parts (ASSPs), system on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0281] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. Such program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0282] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0283] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a cathode ray tube or liquid crystal display, monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0284] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include local area networks, wide area networks, and the Internet.

[0285] A computer system may include a client device and a server. The client device and server are generally remote from each other and typically interact via a communication network. The client device and server relationship arises through computer programs running on the respective computers and having a client device-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0286] It should be noted that the serial numbers of the above-mentioned embodiments of the present disclosure are only for description and do not represent the advantages or disadvantages of the embodiments.

[0287] In the above embodiments of the present disclosure, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0288] In the several embodiments provided in this disclosure, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0289] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0290] In addition, the functional units in the various embodiments of the present disclosure may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0291] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present disclosure. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a read-only memory random access memory, a mobile hard disk, a magnetic disk or an optical disk.

[0292] The above is only a preferred embodiment of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present disclosure. These improvements and modifications should also be regarded as within the scope of protection of the present disclosure.

Claims

1. An information inference method, wherein, Including: Detecting query information to be inferred; Decomposing and processing the query information to obtain sub-query information at different inference stages, where there is a logical association relationship between the sub-query information at different inference stages; Determining sub-response information respectively matching the sub-query information at different inference stages; Verifying the sub-response information matching the sub-query information to obtain a verification result corresponding to the sub-query information, where the verification result is used to characterize the feasibility degree of the sub-query information for inferring the query information; In response to the verification result being greater than the verification result threshold, inferring the sub-query information corresponding to the verification result to obtain a response information corresponding to the query information.

2. The method according to claim 1, wherein Decomposing and processing the query information to obtain sub-query information at different inference stages, including: Based on the query information, searching for the sub-query information at different inference stages in a sample set, where the sample set includes the sub-query information corresponding to different query information.

3. The method according to claim 2, wherein, Based on the query information, searching for the sub-query information at different inference stages in a sample set, including: Determining the identification information of the query information; In the sample set corresponding to the inference stage, determining the sub-query information whose identification information matches the identification information of the query information as the sub-query information of the inference stage.

4. The method according to claim 3, wherein, In the sample set corresponding to the inference stage, determining the sub-query information whose identification information matches the identification information of the query information as the sub-query information of the inference stage, including: In the sample set corresponding to the inference stage, retrieving sub-query information whose similarity between the identification information and the identification information of the query information is greater than the similarity threshold; Determining the retrieved sub-query information as the sub-query information of the inference stage.

5. The method according to claim 3, wherein, Determining the identification information of the query information, including: Using a first large model to analyze the query information to obtain the identification information of the query information.

6. The method according to claim 5, wherein, In the sample set corresponding to the inference stage, determining the sub-query information whose identification information matches the identification information of the query information as the sub-query information of the inference stage, including: Using the first large model to analyze the sample set corresponding to the inference stage and the identification information of the query information to obtain sub-query information whose identification information matches the identification information of the query information; Using the first large model to determine the matched sub-query information as the sub-query information of the inference stage.

7. The method according to claim 5, wherein, Determining sub-response information respectively matching the sub-query information at different inference stages, including: Using the first large model to analyze the sub-query information at different inference stages to obtain sub-response information matching the sub-query information at different inference stages.

8. The method according to claim 1, wherein, Verifying the sub-response information matching the sub-query information to obtain a verification result corresponding to the sub-query information, including: When running to the current inference stage among the different inference stages, verifying the sub-response information matching the sub-query information of the current inference stage to obtain the sub-verification result of the current inference stage, where the verification result includes the sub-verification result; or, Verifying the sub-response information of the different inference stages to obtain the overall verification result of the different inference stages, where the verification result includes the overall verification result.

9. The method according to claim 8, wherein, Verifying the sub-response information of the current inference stage among the different inference stages to obtain the sub-verification result of the current inference stage, including: Using the second largest model to verify the sub-response information of the current inference stage to obtain the sub-verification result; Verifying the sub-response information of the different inference stages to obtain the overall verification result of the different inference stages, including: using the second largest model to verify the sub-response information of the different inference stages to obtain the overall verification result.

10. The method according to claim 8, wherein, Decomposing the query information to obtain sub-query information of different inference stages, including: In the sample set corresponding to the current inference stage, searching for the sub-query information of the current inference stage based on the query information; When the next inference stage of the current inference stage is included in the different inference stages, in the sample set corresponding to the next inference stage, searching for the sub-query information of the next inference stage based on the sub-verification result of the current inference stage and the query information.

11. The method according to any one of claims 1 to 10, wherein, The method further includes: Determining the query information as the root node and the sub-query information as the leaf node to establish a target tree structure; Inferring the response information corresponding to the query information from the sub-query information corresponding to the verification result, including: in the target tree structure, inferring the response information corresponding to the query information from the sub-query information corresponding to the verification result.

12. An information inference method, wherein, Including: In response to the query information to be inferred received on the dialogue interface, on the dialogue interface, displaying the sub-query information of the query information in different inference stages, where there is a logical association relationship between the sub-query information of the different inference stages; On the dialogue interface, displaying the sub-response information respectively matching the sub-query information of the different inference stages; On the dialogue interface, displaying the response information corresponding to the query information, where the response information is obtained by inferring the sub-query information corresponding to the verification result when the verification result corresponding to the sub-query information is greater than the verification result threshold, the verification result is obtained by verifying the sub-response information matching the sub-query information, and is used to characterize the feasibility degree of the sub-query information for inferring the query information.

13. The method according to claim 12, wherein The query information is multi-modal information, and the types of the multi-modal information include at least one of the following: text information including character information, video frame information including frame image information, audio information, and the types of the response information include at least one of the following: text information, image information, video information, and voice information.

14. A method for generating a model, wherein, Including: Obtaining a query information sample; Training a first large model using the sub-query information samples corresponding to the query information samples, where the first large model is used to decompose the input query information to obtain sub-query information at different inference stages, and there is a logical association relationship between the sub-query information at different inference stages; Obtaining the reply information samples corresponding to the query information samples; Obtaining the verification result samples corresponding to the reply information samples, where the verification result samples are used to at least represent the classification results of the reply information samples; Training a second large model using the verification result samples, where the second large model is used to verify the sub-reply information matched by the sub-query information to obtain the verification result corresponding to the sub-query information, and the verification result is used to characterize the feasibility of the sub-query information for inferring the query information, and the sub-query information corresponding to the verification result greater than the verification result threshold is used to infer the reply information corresponding to the query information.

15. The method according to claim 14, wherein, Training a first large model using the sub-query information samples corresponding to the query information samples includes: Determining the loss degree of the sub-query information samples; Training a generation model using the loss degree and a sample set to obtain the first large model, where the sample set includes sub-query information corresponding to different query information, and the generation model is a pre-trained generative large model.

16. The method according to claim 14, wherein, Training a second large model using the verification result samples includes: Determining the loss degree of the verification result samples; Training a generation model using the loss degree of the verification result samples to obtain the second large model, where the generation model is a pre-trained generative large model.

17. An information inference system, wherein, Including: An information generation end for detecting a query information to be inferred, decomposing the query information to obtain sub-query information at different inference stages, and determining sub-reply information respectively matched with the sub-query information at different inference stages, where there is a logical association relationship between the sub-query information at different inference stages; An information verification end for verifying the sub-reply information matched by the sub-query information to obtain the verification result corresponding to the sub-query information, and in response to the verification result being greater than the verification result threshold, inferring the sub-query information corresponding to the verification result to obtain the reply information corresponding to the query information, where the verification result is used to characterize the feasibility of the sub-query information for inferring the query information.

18. An electronic device, wherein, Including: A memory and a processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, and when the computer-executable instructions are executed by the processor, the steps of the method described in any one of claims 1 to 16 are implemented.

19. A computer-readable storage medium, wherein, A computer program is stored, and when the computer program is executed by a computer, the steps of the method described in any one of claims 1 to 16 are implemented.

20. A computer program product comprising a computer program, wherein, When the computer program is executed by the processor, the steps of the method described in any one of claims 1 to 16 are implemented.

Images