Task processing method and device

Abstract

The invention discloses a task processing method and device, and the method comprises the steps: obtaining a target task; determining a plurality of agents with different processing capabilities according to the target task; processing the target task by using a plurality of agents to obtain a plurality of first results, and outputting one first result by each agent; and according to the number of different first results, processing the plurality of first results according to the first processing mode or the second processing mode to obtain a final result of the target task.

Description

Technical Field

[0001] The present application relates to the technical field of task processing, and in particular to a task processing method and device. Background Art

[0002] At present, many electronic devices are locally equipped with data processing models based on artificial intelligence technology. Electronic devices can use these models to analyze and process user input instructions to complete corresponding tasks, such as analyzing specific documents according to user instructions, or generating documents according to user instructions.

[0003] The processing power of a data processing model is positively correlated with the number of parameters in the model, that is, the larger the model is and the more parameters it contains, the stronger the processing power of the model is. However, the more model parameters there are, the more computing resources an electronic device consumes when running the model. Therefore, due to the limitations of the local computing resources of the electronic device, the number of parameters of the model configured locally on the device is limited, and the processing power displayed is sometimes insufficient to meet the needs of users. Summary of the invention

[0004] To this end, this application discloses the following technical solutions:

[0005] The first aspect of the present application provides a task processing method, comprising:

[0006] Obtain the target task;

[0007] Determining a plurality of intelligent agents with different processing capabilities according to the target task;

[0008] Using a plurality of the intelligent agents to process the target task, a plurality of first results are obtained, and each of the intelligent agents outputs a first result;

[0009] According to the number of different first results, the plurality of first results are processed in a first processing mode or a second processing mode to obtain a final result of the target task.

[0010] Optionally, the processing of the plurality of first results in a first processing mode or a second processing mode according to the number of different first results to obtain a final result of the target task includes:

[0011] The step of processing the first results in the first processing mode or the second processing mode is performed multiple times according to the number of different first results until a final result of the target task that meets the convergence condition is obtained.

[0012] Optionally, according to the number of different first results, processing the plurality of first results in a first processing mode or a second processing mode to obtain a final result of the target task includes:

[0013] In the case where there are two different first results, the plurality of first results are processed according to a first processing mode to obtain a final result of the target task;

[0014] In the case where there are more than two different first results, the plurality of first results are processed in a second processing mode to obtain a final result of the target task.

[0015] Optionally, the processing of the plurality of first results in the first processing mode to obtain a final result of the target task includes:

[0016] Using the first agent to process the first result output by the second agent to obtain the secondary task of the first agent, and using the second agent to process the first result output by the first agent to obtain the secondary task of the second agent, wherein the first agent and the second agent are any two agents whose first results are different;

[0017] Using the second agent to process the secondary task of the first agent to obtain a secondary result of the second agent;

[0018] Processing the secondary result of the second agent and a plurality of the first results by using the first agent to obtain a second result of the first agent;

[0019] Using the first agent to process the secondary task of the second agent to obtain a secondary result of the first agent;

[0020] Processing the secondary result of the first agent and the plurality of the first results by the second agent to obtain a second result of the second agent;

[0021] A final result of the target task is determined according to the second result of the first agent and the second result of the second agent.

[0022] Optionally, processing the plurality of first results in the second processing mode to obtain a final result of the target task includes:

[0023] For each of the intelligent agents, use the intelligent agent to process a plurality of the first results to obtain a second result of the intelligent agent;

[0024] A final result of the target task is determined according to the plurality of second results.

[0025] Optionally, also include:

[0026] Obtaining a benchmark result corresponding to the target task;

[0027] Guidance information is determined according to the benchmark result and the final result, and the guidance information is used to instruct the agent to process a next target task.

[0028] Optionally, it also includes;

[0029] Classifying the plurality of guidance information into a plurality of categories;

[0030] The guidance information of each category is processed separately to obtain a guidance information summary corresponding to each category, wherein the guidance information summary is used to instruct the agent to process the next target task belonging to the corresponding category.

[0031] Optionally, the determining of a plurality of agents with different processing capabilities according to the target task includes at least one of the following:

[0032] Divide the target task into a plurality of subtasks, and determine a plurality of agents whose corresponding processing capabilities match the subtasks;

[0033] Determine a plurality of roles related to the target task, and determine a plurality of agents whose corresponding processing capabilities match the roles.

[0034] Optionally, the using a plurality of the intelligent agents to process the target task to obtain a plurality of first results includes:

[0035] Obtaining task instruction information corresponding to the target task;

[0036] The task instruction information and the task data of the target task are input into each of the intelligent agents to obtain a first result output by each of the intelligent agents.

[0037] A second aspect of the present application provides a task processing device, comprising:

[0038] An acquisition unit, used to obtain a target task;

[0039] A determination unit, configured to determine a plurality of intelligent agents with different processing capabilities according to the target task;

[0040] A processing unit for:

[0041] Using a plurality of the intelligent agents to process the target task, a plurality of first results are obtained, and each of the intelligent agents outputs a first result;

[0042] According to the number of different first results, the plurality of first results are processed in a first processing mode or a second processing mode to obtain a final result of the target task. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are merely embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0044] Figure 1 is a flowchart of a task processing method provided by an embodiment of the present application;

[0045] Figure 2 is a flowchart of another task processing method provided by an embodiment of the present application;

[0046] Figure 3 is a schematic diagram of processing a first result based on a first processing mode provided in an embodiment of the present application;

[0047] Figure 4 is a schematic diagram of processing a first result based on a second processing mode provided in an embodiment of the present application;

[0048] Figure 5 is an example diagram of a task processing method provided in an embodiment of the present application;

[0049] Figure 6 It is a structural diagram of a task processing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0051] This application embodiment provides a task processing method, see Figure 1 , the method may include the following steps.

[0052] S101, obtain the target task.

[0053] The target task can be any task that needs to be processed using a target model, for example, a long text reading task, a knowledge extraction task, a product development task, etc. The target model here can be a neural network model, a deep learning model, or other similar models based on artificial intelligence technology. The structure and principle of the target model can refer to the relevant technologies in the field of artificial intelligence, without limitation.

[0054] The target task obtained may include the task data required to process the target task. For example, for a long text reading task, the corresponding task data may include the long text data to be read; for a knowledge extraction task, the task data may include several text documents from which knowledge needs to be extracted; for a product development task, the task data may include data describing product requirements and product application scenarios.

[0055] When executing the method of this embodiment, the electronic device can obtain the target task according to the user's operation, for example, download or read long text data from a storage medium according to the user's operation, and obtain a long text reading task based on the long text data.

[0056] S102, determining a plurality of intelligent agents with different processing capabilities according to the target task.

[0057] In order to process the target task from multiple different angles and obtain processing results of the target task from different angles, in S102, multiple intelligent agents with different processing capabilities and each processing capability is related to the target task may be determined.

[0058] The electronic device may determine the above-mentioned multiple agents in a variety of ways.

[0059] A first optional way to determine the intelligent agent is to divide the target task into multiple subtasks and determine multiple intelligent agents whose corresponding processing capabilities match the subtasks.

[0060] In the first determination method, the electronic device can use the target model to decompose the target task into several subtasks.

[0061] For example, if the target task is a long text reading task, the electronic device can use the target model to decompose the long text reading task into multiple subtasks such as detail analysis task, main idea extraction task, information reasoning task, fact judgment task, problem solving task, etc.

[0062] After dividing the subtasks, for each subtask, the electronic device uses the target model to determine the processing capability required to process the subtask, thereby determining an intelligent agent with the processing capability.

[0063] Combined with the above example, the target model can determine that the processing power required for the detail analysis task is to identify text details related to the specific problem to be solved in the long text data, such as phrases and sentences related to the problem;

[0064] For information reasoning tasks, the target model can determine the required processing power to logically reason about the existing information in the long text data to obtain new information required to solve a specific problem;

[0065] For fact judgment tasks, the target model can determine the required processing power to combine known information and inferred information to determine whether the information recorded in the long text data is correct;

[0066] For the main idea extraction task, the target model can determine the required processing power to summarize and generalize the long text data and determine the theme of the long text data;

[0067] For problem-solving tasks, the target model can be used to determine the required processing power to determine solutions or suggestions related to the problem to be solved based on the content of the long text data.

[0068] After determining the processing power required for each subtask, the electronic device can determine the intelligent agent corresponding to each subtask based on the processing power.

[0069] A second alternative way to identify the agent.

[0070] Identify multiple roles related to the target task and determine multiple agents with corresponding processing capabilities and role matching.

[0071] In the second determination method, the electronic device can use the target model to analyze which different roles of people can handle the target task when the target task is manually processed, and then determine the processing capabilities corresponding to each role, and finally determine multiple intelligent agents with processing capabilities that match each role.

[0072] As an example, if the target task is a product development task, the target model can be used to determine that the target task requires personnel in multiple roles such as product manager, designer, development engineer, test engineer, product operation, etc. to handle it, and then the processing capability of the product manager role is determined based on the target model. Thus, an intelligent agent corresponding to the product manager role and having the processing capability of the product manager role is determined. Similarly, the intelligent agent corresponding to the development engineer role, the intelligent agent corresponding to the test engineer role, and the intelligent agent corresponding to the product operation role can be determined based on the target model.

[0073] As another example, if the target task is a long text reading task, the writer role, editor role, and reviewer role related to the task can be determined, thereby determining the intelligent agent corresponding to the writer role, the intelligent agent corresponding to the editor role, and the intelligent agent corresponding to the reviewer role.

[0074] The electronic device may also combine multiple methods to jointly determine multiple intelligent agents. For example, three intelligent agents with different processing capabilities may be determined according to a first determination method, and three intelligent agents with different processing capabilities may be determined according to a second determination method. Subsequent steps may then be performed based on these six intelligent agents, each of which has different processing capabilities.

[0075] In some embodiments, the electronic device may also determine the multiple agents required in S102 based on the user's operation. For example, the user may input descriptive information describing the processing capabilities of each agent. The electronic device understands the descriptive information based on the target model, and then determines the agent with the corresponding processing capability based on the descriptive information. Alternatively, the electronic device may display a variety of commonly used agents with different processing capabilities, and the user may select multiple agents from them. The electronic device determines the multiple agents selected by the user as the multiple agents used to process the target task in S102.

[0076] S103, using multiple agents to process the target task to obtain multiple first results, and each agent outputs a first result.

[0077] In this embodiment, the processing capability of each agent can be provided by a target model deployed locally on the electronic device. After multiple agents with different processing capabilities are determined in S102, the target model can be used to generate capability description information corresponding to each agent, and the capability description information represents the processing capability of the corresponding agent.

[0078] In combination with the above example, for agent 1 corresponding to the information reasoning task, the capability description information of agent 1 may include "the agent is able to perform logical reasoning on the existing information in the long text data to obtain new information required to solve specific problems."

[0079] For the agent 2 corresponding to the fact judgment task, the capability description information of the agent 2 may include "the agent can combine known information and inferred information to judge whether the information recorded in the long text data is correct."

[0080] When any agent is needed to process a target task, the capability description information corresponding to the agent can be input into the target model as a prompt, so that the target model processes the target task according to the processing capability corresponding to the agent, thereby obtaining the first result output by the agent.

[0081] Continuing with the above example, when the agent 1 corresponding to the information reasoning task is used to process the target task, the capability description information corresponding to the agent and the task data of the target task can be input into the target model. The target model processes the task data according to the capability of obtaining new information by logical reasoning based on the prompt of the capability description information, and obtains the first result output by the agent 1.

[0082] When the intelligent agent 2 corresponding to the fact judgment task is used to process the target task, the capability description information corresponding to the intelligent agent and the task data of the target task can be input into the target model. The target model processes the task data based on the prompt of the capability description information and the ability to judge the authenticity of the information contained in the task data to obtain the first result output by the intelligent agent 2.

[0083] S104, processing the plurality of first results in a first processing mode or a second processing mode according to the number of different first results to obtain a final result of the target task.

[0084] The first results output by multiple agents may be the same or different. For example, assuming that four first results output by four agents are obtained in S103, it is possible that the first result of agent 1 is the same as the second result of agent 2, the first results of agent 3 and agent 4 are the same, the first result of agent 1 is different from the first result of agent 3, or the first results of any two agents from agent 1 to agent 4 are different.

[0085] In S104, the electronic device can compare the multiple first results obtained to determine how many different first results are output by the above-mentioned multiple agents respectively, and based on the number of different first results, use different processing modes to process these first results again to obtain the final result of the target task.

[0086] For example, if after comparison it is determined that there are only two different first results in S103, the electronic device can process the multiple first results in the first processing mode to obtain the final result of the target task; if after comparison it is determined that three or more different first results are obtained in S103, the electronic device can process the multiple first results in the second processing mode to obtain the final result of the target task.

[0087] The electronic device can directly compare whether the contents of the first results output by the two intelligent entities are consistent. If the contents are completely consistent, it is determined that the two first results are the same first result. If the contents are not completely consistent, it is determined that the two first results are two different first results.

[0088] The electronic device can also determine the similarity of the first results output by each two intelligent agents based on the target model. If the similarity between the two first results is greater than a specific similarity threshold, it is determined that the two first results belong to the same first result. If the similarity between the two first results is not greater than the similarity threshold, it is determined that the two first results belong to two different first results.

[0089] The beneficial effects of this embodiment are:

[0090] When processing the target task, multiple agents with different processing capabilities can be used to process the target task respectively, and then different processing models are used to further process the multiple first results according to the number of different first results output by the multiple agents, so as to obtain the final result of the target task by combining the different processing capabilities of the multiple agents. Therefore, this solution can process the target task from multiple angles by combining different processing capabilities, and obtain more accurate processing results without increasing the model parameters of the used model.

[0091] In some optional embodiments, the electronic device may repeatedly execute the step of processing multiple first results in the first processing mode or the second processing mode according to the number of different first results until a final result of the target task that meets the convergence condition is obtained.

[0092] See also Figure 2 In this embodiment, the electronic device may first execute step S201 to obtain multiple first results output by multiple agents.

[0093] S201, using multiple agents to process a target task and obtain multiple first results.

[0094] After obtaining a plurality of first results, the electronic device may execute step S202 to determine whether the plurality of first results meet a convergence condition.

[0095] If the multiple first results do not meet the convergence condition, step S203 may be executed; if the multiple first results meet the convergence condition, step S204 may be executed.

[0096] The convergence condition can be set as needed. For example, the convergence condition can be set to that the multiple first results currently obtained belong to the same first result, that is, assuming that three first results are obtained in S201, if each of the first results is the same as or highly similar to the other two first results, it can be determined that the three first results belong to the same first result and meet the convergence condition; if each of the first results is different from the other two first results or has a low similarity, it can be determined that the three first results do not belong to the same first result and do not meet the convergence condition.

[0097] Alternatively, the convergence condition may be set to that the number of iterations is greater than or equal to a set iteration number threshold. In this embodiment, the number of iterations may be defined as the number of times step S203 is executed.

[0098] Optionally, the convergence condition may also include the above two conditions, that is, as long as any one of the above two conditions is met, it can be determined that the multiple first results obtained meet the convergence condition, and step S204 is executed. If both of the above two conditions are not met, it can be determined that the multiple first results obtained do not meet the convergence condition, and step S203 is executed.

[0099] S203: Process the plurality of first results in the first processing mode or the second processing mode according to the number of different first results.

[0100] In step S203, the electronic device can select the first processing mode or the second processing mode according to the number of different first results, and then based on the selected processing mode, again use the aforementioned multiple agents to process the currently obtained first results to obtain multiple results output by multiple agents.

[0101] Among them, if the number of different first results is equal to 2, the electronic device can select the first processing mode, and use the multiple first results obtained by processing the aforementioned multiple intelligent agents in the first processing mode; if the number of different first results is greater than 2, the electronic device can select the second processing mode, and use the multiple first results obtained by processing the aforementioned multiple intelligent agents in the second processing mode.

[0102] After executing S203, the electronic device can obtain multiple results output by multiple agents. At this time, the electronic device can use these results as new first results and return to execute step S202. If these new first results still do not meet the convergence conditions, the electronic device can execute S203 again, and so on.

[0103] Exemplarily, after the electronic device executes S201, it obtains multiple first results of the first batch, and determines that the multiple first results of the first batch do not meet the convergence condition, so the electronic device executes S203 for the first time, processes the first batch of first results based on multiple agents in the corresponding processing mode, and obtains multiple first results of the second batch. After judgment in S202, it is determined that the first results of the second batch also do not meet the convergence condition, so the electronic device executes S203 for the second time, processes the second batch of first results based on multiple agents in the corresponding processing mode, and obtains multiple first results of the third batch. Assuming that the multiple first results of the third batch meet the convergence condition, the electronic device can execute S204 to determine the final result of the target task according to the multiple first results of the third batch.

[0104] S204, obtaining a final result of the target task that meets the convergence condition.

[0105] In step S204, if the same multiple results are obtained through one or more processes of step S203, any one of them can be selected as the final result of the target task.

[0106] If two or more different results are still obtained through one or more processes of step S203, the electronic device can evaluate the accuracy of each result according to the target model and take the result with higher accuracy than the other results as the final result.

[0107] Alternatively, the electronic device can determine the number of votes for each result and determine the result with the highest number of votes as the final result. The number of votes for a result can be understood as the number of intelligent agents that output this result. For example, assuming there are 5 intelligent agents, 3 of which output the same first result, then the number of votes for this first result is 3.

[0108] The beneficial effects of this embodiment are:

[0109] Performing multiple iterative processing on the results output by multiple agents is beneficial to improving the accuracy of the final results obtained.

[0110] In this embodiment, the first processing mode can be regarded as a confrontation mode between multiple agents. When processing multiple first results according to the first processing mode, the electronic device can use multiple agents to confront each other to obtain the final result of the target task.

[0111] Combine the following Figure 3 The process of processing a plurality of first results based on a first processing mode is described.

[0112] like Figure 3 As shown, in the first processing mode, any two agents with different first output results can be randomly selected as the first agent and the second agent.

[0113] For example, assume that there are 5 agents, namely agent 1 to agent 5. Assuming that the first results output by agents 1 to 3 belong to the same result, and the first results output by agents 4 and 5 belong to the same result and are different from the results of agents 1 to 3, then agent 1 can be selected as the first agent and agent 4 as the second agent.

[0114] After determining the first agent and the second agent, Figure 3 As shown, the two agents can be used to process each other's first results respectively to obtain corresponding secondary tasks. In other words, the first agent can be used to process the first result output by the second agent (for example Figure 3 The first result 2) is obtained, and the secondary task of the first agent is obtained (e.g. Figure 3 Secondary task 1), using the second agent to process the first result output by the first agent (e.g. Figure 3 The first result 1) is used to obtain the secondary task of the second agent (e.g. Figure 3 Secondary task 2).

[0115] When the first agent and the second agent are used to process each other's first result, specific prompt information can be input to the two agents so that the agents output specific secondary tasks.

[0116] For example, the input prompt information may be prompt information instructing the agent to question the input first result. Through this prompt information, the first agent may output sub-task 1 questioning the correctness of the first result 2, and the second agent may output sub-task 2 questioning the correctness of the first result 1.

[0117] The content of the prompt information is not limited. As an example, the prompt information instructing the agent to question the first input result may include the following content:

[0118] Challenge_prompt=""Opponent has come up with different answer.Pleasemaintain your perspective and find evidence from the document to challengeand question the opponent,output starts with[Challenge].

[0119] [Opponent's answer]

[0120] {opponent}

[0121] Output uses the following format:

[0122] [Challenge]

[0123] Your questions to opponentbased on the document"".

[0124] After obtaining the secondary tasks output by the two agents, the two agents can be used to process the secondary tasks output by each other respectively to obtain the secondary results corresponding to the secondary tasks.

[0125] For example, Figure 3 In the example, the second agent can be used to process the secondary task of the first agent (i.e., secondary task 1) to obtain the secondary result of the second agent (i.e., secondary result 1);

[0126] The first agent is used to process the secondary task of the second agent (i.e., secondary task 2) to obtain the secondary result of the first agent (i.e., secondary result 2).

[0127] Each agent can process the input sub-task according to its output results and the task data of the target task. In combination with the above example, the first agent can process sub-task 2 according to the first result 1 and the task data of the target task to obtain the corresponding sub-result 2, and the second agent can process sub-task 1 according to the first result 2 and the task data of the target task to obtain the corresponding sub-result 1.

[0128] When processing a secondary task, prompt information for indicating a processing method of the secondary task may be input to the agent, so that the agent processes the input secondary task according to the specified processing method.

[0129] For example, when the first agent is used to process the secondary task 2 of the second agent, the input prompt information can instruct the first agent to answer the questions in the secondary task 2 to illustrate the correctness of the first result 1 of the first agent. When the second agent is used to process the secondary task 1 of the second agent, the input prompt information can instruct the second agent to answer the questions in the secondary task 1 to illustrate the correctness of the first result 2 of the second agent.

[0130] The content of the prompt information is not limited. As an example, when using an intelligent agent to process a secondary task, the following prompt information can be input to the intelligent agent so that the intelligent agent can answer the questions in the secondary task.

[0131] rebute prompt=""Opponent has raised doubts about you. Please maintain your own viewpoint and find evidence from the document to refute theOpponent'sviewpoint,output starts with [Response].

[0132] [Response]

[0133] [Opponent's challenge]

[0134] [opponent]

[0135] Output uses the following format

[0136] [Response]

[0137] Your response to opponent's challenge based on the document"".

[0138] After obtaining the secondary results, the first agent can be used to process the secondary results of the second agent (i.e. Figure 3 The secondary result 1) and multiple first results of the first agent are used to obtain the second result of the first agent, and the secondary result of the first agent is processed by the second agent (i.e. Figure 3 The secondary result 2) and multiple first results are used to obtain the second result of the second agent.

[0139] The first agent and the second agent can combine the secondary results output by each other and the first results of both parties, and process the target task again to obtain their respective second results.

[0140] Optionally, when the two agents process in the above manner to obtain their respective second results, they may also refer to other interaction contents between the first agent and the second agent.

[0141] For example, the first agent can process the target task again to obtain the second result based on the secondary result 1 of the second agent and the first results of both parties. It can also process the target task again to obtain the second result based on multiple contents such as secondary task 1 output by the first agent, secondary task 2 output by the second agent, secondary result 2 output by the first agent, secondary result 1 of the second agent and the first results of both parties.

[0142] When the agent is used to obtain the corresponding second result, specific prompt information may also be input to the agent to instruct the agent to process the target task again based on the interactive content. As an example, the prompt information input to the agent in this case may include the following content.

[0143] reanswer_prompt=""Please read the document again,consider itcomprehensively,consider the previous discussion comprehensively,and come up with your final answer,starts with[Answer],and explaining the reasons,startswith[Reasoning]

[0144] Output uses the following format:

[0145] [Answer]

[0146] Answer number(Select one option).

[0147] [Reasoning]

[0148] Find the most relevant content to answer the question,explain thereasoning porcess,and state the reason for choosing that answer.

[0149] After obtaining the second results of the two agents, the final result of the target task can be determined according to the second result of the first agent and the second result of the second agent.

[0150] After obtaining the second result, if Figure 2 As shown, it can be determined whether the two second results meet the convergence condition. If the convergence condition is met, the final result of the target task can be determined in the manner of step S204. If the convergence condition is not met, the two second results can be determined as a new batch of first results, and then the new batch of first results can be processed again according to the above-mentioned first processing mode until a result that meets the convergence condition is obtained after a certain processing.

[0151] In this embodiment, the second processing mode can be regarded as a cooperation mode between multiple agents. When processing multiple first results according to the second processing mode, the electronic device can process the target task through the cooperation of multiple agents to obtain the final result of the target task.

[0152] Combine the following Figure 4 The process of processing a plurality of first results based on the second processing mode is described.

[0153] When processing a plurality of first results in the second processing mode, each first result may be input into each agent, so that each agent is used to process the plurality of first results to obtain the second result of the agent.

[0154] by Figure 4 For example, assume that three agents are determined based on the target task, namely the first agent, the second agent and the third agent. After processing the target task, the three agents output three different first results, namely the first result 1 to the first result 3. Since the number of different first results is greater than 2, they can be processed according to the second processing mode.

[0155] Therefore, the first results 1 to the first results 3 can be input into the first intelligent agent together, so that the first intelligent agent processes the target task again according to the first results 1 to the first results 3 to obtain the second result 1; the first results 1 to the first results 3 can be input into the second intelligent agent together, so that the second intelligent agent processes the target task again according to these first results to obtain the second result 2; the first results 1 to the first results 3 can be input into the third intelligent agent together, so that the third intelligent agent processes the target task again according to these first results to obtain the second result 3.

[0156] When multiple first results are input into an agent, specific prompt information may be input into the agent to instruct the agent to process the target task again based on the input first results. As an example, the prompt information input into the agent in this case may include the following content.

[0157] cooperate_prompt=""{other_answer}

[0158] Firstly,Based on the answer and reasons of the above people,checkwhether there are and omissions or error in your answer and reasons,checkwhether there are any error in the reasons and answer ofothers,and your demonstration needs to be supported by the article.Output starts with [Demonstration].

[0159] Finally,Answer again use answer number.Output starts with[Answer].Forexample:[Answer]3””.

[0160] After obtaining the plurality of second results, the final result of the target task may be determined according to the plurality of second results.

[0161] The method of determining the final result in the first processing mode is consistent with that in the first processing mode. After obtaining multiple second results in the second processing mode, the final result can also be determined in the second processing mode. Figure 2In step S202, it is determined whether the multiple second results obtained meet the convergence condition. If the convergence condition is met, the final result of the target task can be determined in the manner of step S204. If the convergence condition is not met, the multiple second results obtained can be determined as a new batch of first results, and then S203 is executed again to process the new batch of first results again according to the first processing mode or the second processing mode until a result that meets the convergence condition is obtained after a certain processing.

[0162] In some optional embodiments, after obtaining the final result, the guidance information corresponding to the target task may be determined in the following manner to improve the accuracy of processing other subsequent tasks:

[0163] Obtain benchmark results corresponding to the target task;

[0164] The guidance information is determined based on the baseline results and the final results, and is used to instruct the agent to process the next target task.

[0165] Benchmark results can be obtained in a variety of ways.

[0166] One way to obtain the benchmark result may be to display the final result and the results output by each agent before the final result is obtained to the user, and prompt the user to select the result that best meets the user's needs, and determine the result selected by the user as the benchmark result. Alternatively, the user may be prompted to revise the output final result, and the result after the user's revision may be determined as the benchmark result.

[0167] One way to obtain a benchmark result may be to use a large model containing more parameters to process the target task, and determine the result output by the large model as the benchmark result. Exemplarily, assuming that the above-mentioned agents all process the target task based on a target model with fewer model parameters deployed locally on the electronic device, when obtaining the benchmark result, the electronic device can upload the target task to the server, use the large model containing a large number of model parameters deployed by the server to process the target task, and receive the processing result of the large model fed back by the server as the benchmark result.

[0168] After obtaining the baseline results, the baseline results and the final results can be input into each intelligent agent, and the intelligent agent can be used to compare the differences between the two, and determine the guidance information based on the differences between the two.

[0169] Optionally, after obtaining the baseline result, the final result can be identified first. If the similarity between the baseline result and the final result is high, the final result can be considered to be the correct answer. In this case, the guidance information can be determined based on the baseline result and the final result. If the similarity between the baseline result and the final result is low, the final result can be considered to be an incorrect answer. In this case, the intelligent agent can be used to determine the guidance information based on the baseline result and the final result.

[0170] In the case where some guidance information has been obtained, after obtaining the target task, the agent can refer to the obtained guidance information to process the target task.

[0171] One way to process a target task with reference to guidance information may be to input the required guidance information into the agent together with the task data in the form of prompts when using the agent to process the target task, so that the agent processes the target task with reference to the input guidance information.

[0172] One way to process a target task with reference to guidance information may be, after obtaining the final result of the target task according to the method of the aforementioned embodiment, inputting the final result and guidance information into an intelligent agent, and using the guidance information as a basis, using the intelligent agent to correct the obtained final result to obtain a corrected final result. In this case, the electronic device may finally output the corrected final result to the user.

[0173] In some optional embodiments, when there is a large amount of guidance information, in order to improve the efficiency of the agent in processing tasks based on the guidance information, the plurality of guidance information may be summarized as follows:

[0174] Dividing a plurality of guidance information into a plurality of categories;

[0175] The guidance information of each category is processed separately to obtain the guidance information summary corresponding to each category. The guidance information summary is used to instruct the agent to process the next target task belonging to the corresponding category.

[0176] In this embodiment, the guidance information can be divided into categories according to the type of target task when obtaining the guidance information. For example, the guidance information obtained when processing a long text reading task can be divided into one category, the guidance information obtained when processing a knowledge extraction task can be divided into one category, and the guidance information obtained when processing a product development task can be divided into one category.

[0177] After completing the category division, for each category, multiple guidance information belonging to the category can be input into the target model, and the target model can delete redundant content in the guidance information of the category and merge duplicate content to form a summary of the guidance information of the category.

[0178] The above step of summarizing the guidance information may be performed regularly at a certain period, for example, once a week, once every 10 days, etc.

[0179] When processing any target task, it can be determined first whether there is a guidance information summary corresponding to the category. If there is a guidance information summary of the category, the agent can refer to the guidance information summary of the category to process the target task.

[0180] For example, if a long text reading task that needs to be processed is obtained, it can be determined whether there is a guidance information summary corresponding to the long text reading task. If there is a guidance information summary corresponding to the long text reading task, the intelligent agent can refer to the guidance information summary to process the long text reading task obtained this time.

[0181] The method of processing the target task with reference to the guidance information summary may be consistent with the method of processing the target task with reference to the guidance information described above, and will not be described in detail.

[0182] In some optional embodiments, when multiple agents are used to process a target task and multiple first results are obtained, specific prompt information may be input to the agents in the following manner to improve the processing efficiency of the task:

[0183] Obtaining task instruction information corresponding to the target task;

[0184] Task instruction information and task data of the target task are input into each intelligent agent to obtain a first result output by each intelligent agent.

[0185] The task instruction information may include the guidance information or the guidance information summary of the aforementioned embodiment.

[0186] The task instruction information may also be input by the user when obtaining the target task. In this case, the task instruction information may indicate a processing method of the target task.

[0187] The task processing method of this embodiment can be used to process various tasks. As an example, it can be used to process a knowledge extraction task.

[0188] See also Figure 5 When processing the knowledge extraction task, we can first obtain the target document from which the knowledge needs to be extracted as the task data, and then determine four agents with different processing capabilities, namely agent 1 to agent 4.

[0189] Use Agents 1 to 4 to perform knowledge extraction tasks on the target document respectively, and obtain the first results of each agent, such as Figure 5 As shown, the first results of agent 1 to agent 4 are result 1 to result 4 respectively.

[0190] After comparing the above first results, it is determined that results 1 to 4 are different from each other, that is, there are 4 different first results, so results 1 to 4 can be processed according to the second processing mode.

[0191] When processing in the second processing mode, results 1 to 4 can be input to agents 1 to 4 respectively, so that each of agents 1 to 4 processes the target task again according to results 1 to 4.

[0192] After being processed in the second processing mode, the four agents respectively output two different second results, wherein the second results output by agent 1 and agent 2 are both result 5, and the second results output by agent 3 and agent 3 are both result 6.

[0193] Since there are two different results, we can regard results 5 and 6 as the two first results of the second batch, and determine agent 1 as Figure 3 The first agent shown, agent 4 is determined as Figure 3 The second agent shown in the figure processes results 5 and 6 according to the first processing mode mentioned above. At this time, result 5 is equivalent to Figure 3 The first result is 1, and the result is 6, which is equivalent to Figure 3 The first result is 2.

[0194] After processing according to the first processing mode, the second result output by the first agent is the same as the second result output by the second agent, so it is determined that the second result meets the convergence condition, and the obtained second result is determined as the final result. Then, the first agent or the second agent is used to correct the final result according to the pre-obtained guidance information to obtain the corrected final result, and the corrected final result is stored in the knowledge base as the knowledge extracted from the target document.

[0195] The present application also provides a task processing device, see Figure 6 , the device may include the following units.

[0196] An obtaining unit 601 is used to obtain a target task;

[0197] A determination unit 602, configured to determine a plurality of agents with different processing capabilities according to a target task;

[0198] The processing unit 603 is used to:

[0199] Using multiple agents to process the target task, multiple first results are obtained, and each agent outputs a first result;

[0200] According to the number of different first results, the multiple first results are processed in the first processing mode or the second processing mode to obtain a final result of the target task.

[0201] Optionally, when the processing unit 603 processes the multiple first results in the first processing mode or the second processing mode according to the number of different first results to obtain the final result of the target task, it can be used to:

[0202] The step of processing the first results in the first processing mode or the second processing mode is performed multiple times according to the number of different first results until a final result of the target task that meets the convergence condition is obtained.

[0203] Optionally, when the processing unit 603 processes the multiple first results in the first processing mode or the second processing mode according to the number of different first results to obtain the final result of the target task, it can be used to:

[0204] In the case where there are two different first results, the plurality of first results are processed according to the first processing mode to obtain a final result of the target task;

[0205] In the case where there are more than two different first results, the multiple first results are processed according to the second processing mode to obtain the final result of the target task.

[0206] Optionally, when the processing unit 603 processes the multiple first results according to the first processing mode to obtain the final result of the target task, it can be used to:

[0207] Using the first agent to process the first result output by the second agent, a secondary task of the first agent is obtained; using the second agent to process the first result output by the first agent, a secondary task of the second agent is obtained; the first agent and the second agent are any two agents that have different first results;

[0208] Using the second agent to process the secondary task of the first agent to obtain the secondary result of the second agent;

[0209] Processing the secondary result of the second agent and the plurality of first results by using the first agent to obtain the second result of the first agent;

[0210] Using the first agent to process the secondary task of the second agent, obtaining the secondary result of the first agent;

[0211] Processing the secondary result of the first agent and the plurality of first results by the second agent to obtain a second result of the second agent;

[0212] A final result of the target task is determined according to the second result of the first agent and the second result of the second agent.

[0213] Optionally, when the processing unit 603 processes the multiple first results according to the second processing mode to obtain the final result of the target task, it can be used to:

[0214] For each intelligent agent, use the intelligent agent to process the plurality of first results to obtain a second result of the intelligent agent;

[0215] A final result of the target task is determined according to the plurality of second results.

[0216] Optionally, the processing unit 603 is further configured to:

[0217] Obtain benchmark results corresponding to the target task;

[0218] The guidance information is determined based on the baseline results and the final results, and is used to instruct the agent to process the next target task.

[0219] Optionally, the processing unit 603 is further configured to:

[0220] Dividing a plurality of guidance information into a plurality of categories;

[0221] The guidance information of each category is processed separately to obtain the guidance information summary corresponding to each category. The guidance information summary is used to instruct the agent to process the next target task belonging to the corresponding category.

[0222] Optionally, when the determining unit 602 determines a plurality of agents with different processing capabilities according to the target task, at least one of the following may be performed:

[0223] Divide the target task into multiple subtasks and determine multiple agents whose processing capabilities match the subtasks;

[0224] Identify multiple roles related to the target task and determine multiple agents with corresponding processing capabilities and role matching.

[0225] Optionally, when the processing unit 603 uses multiple agents to process the target task and obtains multiple first results, it can be used to:

[0226] Obtaining task instruction information corresponding to the target task;

[0227] Task instruction information and task data of the target task are input into each intelligent agent to obtain a first result output by each intelligent agent.

[0228] The working principle of the task processing device provided in the embodiment of the present application can refer to the relevant steps of the task processing method provided in any embodiment of the present application, and will not be repeated here.

[0229] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0230] For the convenience of description, the above system or device is described by dividing it into various modules or units according to its functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0231] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the relevant technology in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0232] Finally, it should be noted that, in this article, relational terms such as first, second, third and fourth are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0233] The above is only a preferred implementation of the present application. 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 application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A task processing method, comprising: Obtain the target task; Determining a plurality of intelligent agents with different processing capabilities according to the target task; Using a plurality of the intelligent agents to process the target task, a plurality of first results are obtained, and each of the intelligent agents outputs a first result; According to the number of different first results, the plurality of first results are processed in a first processing mode or a second processing mode to obtain a final result of the target task.

2. The method according to claim 1, wherein the processing of the plurality of first results in a first processing mode or a second processing mode according to the number of different first results to obtain the final result of the target task comprises: The step of processing the first results in the first processing mode or the second processing mode is performed multiple times according to the number of different first results until a final result of the target task that meets the convergence condition is obtained.

3. The method according to claim 1, processing the plurality of first results in a first processing mode or a second processing mode according to the number of different first results to obtain the final result of the target task, comprising: In the case where there are two different first results, the plurality of first results are processed according to a first processing mode to obtain a final result of the target task; In the case where there are more than two different first results, the plurality of first results are processed in a second processing mode to obtain a final result of the target task.

4. The method according to claim 3, wherein the processing of the plurality of first results in a first processing mode to obtain a final result of the target task comprises: Using the first agent to process the first result output by the second agent to obtain the secondary task of the first agent, and using the second agent to process the first result output by the first agent to obtain the secondary task of the second agent, wherein the first agent and the second agent are any two agents whose first results are different; Using the second agent to process the secondary task of the first agent to obtain a secondary result of the second agent; Processing the secondary result of the second agent and a plurality of the first results by using the first agent to obtain a second result of the first agent; Using the first agent to process the secondary task of the second agent to obtain a secondary result of the first agent; Processing the secondary result of the first agent and the plurality of the first results by the second agent to obtain a second result of the second agent; A final result of the target task is determined according to the second result of the first agent and the second result of the second agent.

5. The method according to claim 3, wherein the processing of the plurality of first results in the second processing mode to obtain the final result of the target task comprises: For each of the intelligent agents, use the intelligent agent to process a plurality of the first results to obtain a second result of the intelligent agent; A final result of the target task is determined according to the plurality of second results.

6. The method according to claim 1, further comprising: Obtaining a benchmark result corresponding to the target task; Guidance information is determined according to the benchmark result and the final result, and the guidance information is used to instruct the agent to process a next target task.

7. The method according to claim 6, further comprising: Classifying the plurality of guidance information into a plurality of categories; The guidance information of each category is processed separately to obtain a guidance information summary corresponding to each category, wherein the guidance information summary is used to instruct the agent to process the next target task belonging to the corresponding category.

8. The method according to claim 1, wherein the step of determining a plurality of agents with different processing capabilities according to the target task comprises at least one of the following: Divide the target task into a plurality of subtasks, and determine a plurality of agents whose corresponding processing capabilities match the subtasks; Determine a plurality of roles related to the target task, and determine a plurality of agents whose corresponding processing capabilities match the roles.

9. The method according to claim 1, wherein the using a plurality of the intelligent agents to process the target task to obtain a plurality of first results comprises: Obtaining task instruction information corresponding to the target task; The task instruction information and the task data of the target task are input into each of the intelligent agents to obtain a first result output by each of the intelligent agents.

10. A task processing device, comprising: An acquisition unit, used to obtain a target task; A determination unit, configured to determine a plurality of intelligent agents with different processing capabilities according to the target task; A processing unit for: Using a plurality of the intelligent agents to process the target task, a plurality of first results are obtained, and each of the intelligent agents outputs a first result; According to the number of different first results, the plurality of first results are processed in a first processing mode or a second processing mode to obtain a final result of the target task.

Images