Artificial intelligence-based task processing method, apparatus and system
By introducing a single-round reasoning and decision-making process into the intelligent agent, the problems of slow response speed and high resource consumption caused by large language models are solved, and rapid response and resource saving are achieved in scenarios such as smart speakers and in-vehicle systems.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI XIAODU TECHNOLOGY CO LTD
- Filing Date
- 2026-01-20
- Publication Date
- 2026-05-29
AI Technical Summary
The standard reasoning process of existing intelligent agents is slow and resource-intensive due to the large number of parameters in the Large Language Model (LLM), making it difficult to meet the needs of scenarios that are sensitive to latency and have strict cost control.
A single-round reasoning and decision-making process is adopted. The task request is pre-processed by the pre-executor to determine whether it meets the target requirements. If it does not meet the requirements, the target agent is called to perform the standard reasoning process. Otherwise, the result of the pre-processing is used as the result of the target processing.
It improves response speed and reduces resource consumption, making it suitable for scenarios that are sensitive to latency and have strict cost control, such as smart speakers and in-vehicle systems.
Smart Images
Figure CN122114145A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the field of artificial intelligence technology, particularly to the fields of large models, intelligent agents, and intelligent interactive systems, and specifically to a task processing method, device, intelligent system, equipment, medium, and product based on artificial intelligence. Background Technology
[0002] With technological advancements, intelligent agents can be used to perform tasks based on reasoning processes. The standard reasoning process for an intelligent agent includes a loop of Thought, Action, and Observation.
[0003] The core component of an intelligent agent is the Large Language Model (LLM). The large number of parameters in an LLM results in a slow response time and high resource consumption during task execution. Summary of the Invention
[0004] This disclosure provides a task processing method, apparatus, system, device, medium, and product based on artificial intelligence.
[0005] According to one aspect of this disclosure, an artificial intelligence-based task processing method is provided, comprising: performing a single-round reasoning process on a task request to obtain a pre-processing result; making a decision on the pre-processing result to obtain a decision result; the decision result being used to indicate whether the pre-processing result meets the target requirements, and a target intelligent agent to be invoked when it does not meet the target requirements; in response to the decision result indicating that the pre-processing result does not meet the target requirements, using the target intelligent agent to perform a standard reasoning process on the task request to obtain the target processing result of the task request.
[0006] According to another aspect of this disclosure, an artificial intelligence-based task processing device is provided, comprising: a first processing module, configured to perform a single-round reasoning process on a task request to obtain a pre-processing result; a decision module, configured to make a decision on the pre-processing result to obtain a decision result; the decision result being used to indicate whether the pre-processing result meets the target requirements, and a target intelligent agent to be invoked when it does not meet the target requirements; and a second processing module, configured to, in response to the decision result indicating that the pre-processing result does not meet the target requirements, use the target intelligent agent to perform a standard reasoning process on the task request to obtain the target processing result of the task request.
[0007] According to another aspect of this disclosure, an artificial intelligence-based task processing system is provided, comprising: a pre-processor for performing a single-round reasoning process on a task request to obtain a pre-processing result; and, when the pre-processing result does not meet the target requirements, determining a target agent; and a target agent for receiving the task request sent by the pre-processor and performing a standard reasoning process on the task request to obtain the target processing result of the task request.
[0008] According to another aspect of this disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to said at least one processor; wherein the memory stores instructions executable by said at least one processor, said instructions being executed by said at least one processor to enable said at least one processor to perform the method as described in any of the foregoing aspects.
[0009] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method according to any of the preceding aspects.
[0010] According to another aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method according to any of the preceding aspects.
[0011] According to embodiments of this disclosure, response speed can be improved and resource consumption can be reduced.
[0012] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description
[0013] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:
[0014] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure;
[0015] Figure 2 This is a schematic diagram of the overall process provided according to the embodiments of this disclosure;
[0016] Figure 3 This is a schematic diagram according to the second embodiment of the present disclosure;
[0017] Figure 4 This is a schematic diagram according to the third embodiment of the present disclosure;
[0018] Figure 5This is a schematic diagram according to the fourth embodiment of the present disclosure;
[0019] Figure 6 This is a schematic diagram of an electronic device used to implement the artificial intelligence-based task processing method of the embodiments of this disclosure. Detailed Implementation
[0020] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.
[0021] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure. This embodiment provides a task processing method based on artificial intelligence. Figure 1 As shown, the method includes:
[0022] 101. Perform a single round of reasoning on the task request to obtain the pre-processing results.
[0023] 102. Make a decision on the preprocessing result to obtain a decision result; the decision result is used to indicate whether the preprocessing result meets the target requirements, and if it does not meet the target requirements, the target intelligent agent to be invoked.
[0024] 103. In response to the decision result indicating that the preprocessing result does not meet the target requirements, the target agent performs a standard reasoning process on the task request to obtain the target processing result of the task request.
[0025] The method in this embodiment can be executed by an artificial intelligence-based task processing system. Specifically, the system can be a conversational artificial intelligence system that has comprehensive requirements for interaction latency, operating costs, and complex task processing capabilities, such as a smart speaker interaction system, an in-vehicle system, and a smart home appliance control system.
[0026] A task request is the input information to the aforementioned system, used to trigger the system to complete a corresponding task. Specifically, a task request can be a question, which the system processes to obtain the answer to. Alternatively, a task request can be a control command, which the system executes a corresponding operation based on.
[0027] In the context of intelligent agents, in related technologies, task requests are directly input into the intelligent agent containing an LLM. The intelligent agent obtains the target processing result corresponding to the task request through a standard loop process of Thought (thought chain reasoning), Action (tool invocation), and Observation (result feedback), such as obtaining the answer to the question or executing the operation corresponding to the control command.
[0028] However, in scenarios such as smart speakers that are sensitive to latency and have strict cost control, it will be difficult to meet the requirements of rapid response if a unified LLM-driven agent is used, while bringing additional and unsustainable costs and response speed burdens.
[0029] Therefore, in this embodiment, task requests are not directly input into the agent. Instead, preprocessing is performed first. Only after determining that an agent is needed is the task request input into the agent, which then executes the standard reasoning process. For some simple tasks, the preprocessing result can be used as the target processing result of the task request, i.e., the final task completion result.
[0030] The pre-processing unit is used to perform the above-mentioned pre-processing, which may include: a single-round reasoning process and a decision-making process.
[0031] A single-round reasoning process includes a single Thought (thought chain reasoning), Action (tool invocation), and Observation (result feedback) process, which can be represented by first-Thought, first-Action, and first-Observation, respectively.
[0032] Through the single-round reasoning process described above, the pre-processing results of the task request are obtained.
[0033] The decision-making process is used to determine whether the results of the preliminary processing meet the target requirements.
[0034] The target requirement is the completion requirement corresponding to the task request. That is, if the target requirement is met, the task is completed; otherwise, the task is not completed.
[0035] When the result of the preprocessing meets the target requirements, the result of the preprocessing is used as the target processing result of the task request. At this time, there is no need to call the intelligent agent.
[0036] If the results of the preprocessing do not meet the target requirements, it indicates that the task has not been completed. At this time, the intelligent agent is invoked to complete the task.
[0037] There can be one or more intelligent agents. If there is only one intelligent agent, then that intelligent agent is used as the target intelligent agent. If there are multiple intelligent agents, then one of the multiple intelligent agents can be determined as the target intelligent agent.
[0038] After the target agent is identified, the preceding executor sends the task request to the target agent. The target agent contains an LLM and performs a standard reasoning process based on the LLM, which is a loop of Thought (thought chain reasoning), Action (tool invocation), and Observation (result feedback) until the target processing result of the task request is obtained.
[0039] In this embodiment, by executing a single-round inference process and executing a standard inference process when the preprocessing result does not meet the target requirements, the first round of inference process can be decoupled from the non-first round of inference process. A decision is also made after the first round of inference process. When the preprocessing result does not meet the target requirements, the target agent executes the standard inference process on the task request, while the non-agent directly executes the standard inference process on the task request. This can minimize unnecessary agent calls, thereby improving response speed and reducing resource consumption.
[0040] Figure 2 This is a schematic diagram of the overall process provided according to the embodiments of this disclosure.
[0041] like Figure 2 As shown, the overall process can be divided into a pre-processing process 201 and a standard reasoning process 202. The pre-processing process includes a single-round reasoning process and a determination process.
[0042] A single-round reasoning process includes a single thought (thought chain reasoning), action (tool invocation), and observation (result feedback), which are represented by the first thought chain reasoning, the first action, and the first observation, respectively.
[0043] The front-end executor receives task requests and, using a small model and pre-set information such as pre-set templates or rules, performs thought chain reasoning to obtain the reasoning result. The intelligent agent includes an LLM (Limited Module Model), but the small model is smaller than the LLM, meaning it has fewer parameters and is less expensive, thus enabling low-cost and low-latency execution of related processing.
[0044] Taking the smart speaker system as an example, the system has accumulated a large amount of information such as templates or rules, which can make full use of these existing resources and achieve effective compatibility with existing resources.
[0045] The reasoning result is obtained through the first thought chain reasoning. This reasoning result may include the name of the tool that needs to be called, such as the weather query application program interface (API), alarm clock setting API, etc.
[0046] The initial action can be based on this reasoning result and perform the corresponding action, such as calling the weather query API.
[0047] After the action is performed, the corresponding observation result is obtained through the first observation, such as obtaining the queried weather data after calling the weather query API.
[0048] After obtaining the observation results, it can be determined whether the target requirements are met, that is, whether the task corresponding to the task request has been completed.
[0049] Specifically, the observation results and related data can be input into a pre-trained classification model. The output of the classification model includes whether the target requirements are met, and if the target requirements are not met, it can also indicate the information of the target agent to be invoked.
[0050] If the number of agents is N (a positive integer), then the classification results of the classification model are (N+1) types, which correspond to whether an agent needs to be called and the specific agent to be called.
[0051] When the target requirements are met, the observation result of the first observation is used as the target processing result. When the target requirements are not met, the task request is sent to the target agent. The target agent executes the standard reasoning process, that is, the loop process of thought chain reasoning, action and observation, to obtain the target processing result.
[0052] In conjunction with the above, this disclosure also provides the following embodiments.
[0053] Figure 3 This is a schematic diagram based on a second embodiment of the present disclosure. This embodiment provides a task processing method based on artificial intelligence. Figure 3 As shown, the method includes:
[0054] 301. Perform a single-round reasoning process on the task request to obtain the pre-processing results.
[0055] Specifically, a small model can be used to perform thought chain reasoning on the task request to obtain a reasoning result; based on the reasoning result and preset information, a target tool is determined; and the target tool is invoked to obtain the preprocessing result.
[0056] Specifically, refer to Figure 2The single-round reasoning process includes the first thought chain reasoning, the first action, and the first observation, which are executed sequentially. The target tool is determined through the first thought chain reasoning, such as obtaining the name of the target tool. The corresponding target tool is called according to the name through the first action. The observation result is obtained through the first observation, which is the execution result of the target tool. The observation result is used as the preprocessing result.
[0057] In this embodiment, the target tool is determined based on the reasoning results and preset information. This method is compatible with existing resources, and combining multiple types of information to determine the target tool can improve accuracy.
[0058] 302. Input the information and associated data of the target tool into a pre-trained classification model to obtain the decision result output by the classification model; the decision result is used to indicate whether the preprocessing result meets the target requirements, and the target agent to be invoked when it does not meet the target requirements.
[0059] This allows you to input information about the target tool, such as its name, and related data into the classification model. The related data is existing data, such as statistical data (e.g., song popularity) and user settings data (e.g., usage time limits).
[0060] The output of the classification model after processing the above information is the classification result, which indicates whether the target requirements are met, and if not, the target agent.
[0061] For example, in a scenario involving an agent, the classification result includes 0 or 1, where 0 indicates that the agent does not need to be invoked (meets the target requirements), and 1 indicates that the agent needs to be invoked.
[0062] In scenarios with multiple agents, such as two agents, if the classification result is 010, corresponding to not calling an agent, calling the first agent, and calling the second agent respectively, then it indicates that the target agent is the first agent.
[0063] In this embodiment, a classification model is used to determine whether the preprocessing results meet the target requirements and the target intelligent agent. This allows for simple and efficient classification, enabling high-precision traffic distribution.
[0064] 303. In response to the decision result indicating that the preprocessing result does not meet the target requirements, the target agent performs a standard reasoning process on the task request to obtain the target processing result of the task request.
[0065] Specifically, the process involves: acquiring historical tools corresponding to the task request; removing preset type tools from the historical tools and the target tools to obtain filtered tools; and sending the filtered tools and the task request to the target agent to enable the target agent to execute a standard reasoning process and obtain the target processing result of the task request.
[0066] Specifically, the interaction process can be recorded to obtain the historical tools corresponding to the task request. The target tool belongs to the current tool. Preset type tools are removed from the historical tools and the current tools to obtain the filtered tools. For example, if the historical tools and the current tools include a first tool, a second tool, and a third tool, and assuming that the second tool is a preset type tool, then the second tool is removed, and the filtered tools include the first tool and the third tool. Then, the target agent performs a standard inference process based on the filtered tools and the task request to obtain the target processing result.
[0067] Preset type tools can specifically be analogous nondeterministic tools. The characteristic of such tools is that they return nondeterministic results. If these results are sent to the agent, it will increase the agent's understanding burden.
[0068] For example, for alarm clock settings, a safe mode can be set. In this mode, the feedback result requires user confirmation. For example, if the user command is "set alarm at 7 o'clock", the tool simulates the interface instead of directly executing the setting operation. The feedback data is such as "confirm setting alarm at 7 o'clock?". This feedback data can prevent user misoperation, but it is difficult for the intelligent agent to understand. That is, based on this feedback data, the intelligent agent has difficulty deciding whether to confirm, provide feedback to the user, or call other interfaces.
[0069] To avoid increasing the decision-making burden on the intelligent agent, the aforementioned preset type tools can be executed in the front-end execution unit according to preset rules or traditional dialogue management methods, instead of being sent to the intelligent agent. For example, the feedback results of the aforementioned tools can be fed back to the user. In this way, the preset type tools can be executed in the front-end execution unit without being sent to the intelligent agent.
[0070] In this embodiment, by removing preset types of tools, tools that are difficult for the agent to understand can be filtered out, reducing the agent's processing burden and improving the accuracy of the processing results. By acquiring historical tools, processing them, and then sending them to the target agent, the target agent can refer to historical data, reducing the illusion problem and improving accuracy.
[0071] 304. In response to the decision result indicating that the preprocessing result meets the target requirements, the preprocessing result is taken as the target processing result of the task request.
[0072] That is, when the preprocessing result meets the target requirements, the target processing result can be directly obtained in the preprocessing stage without calling the agent, which can reduce latency and cost.
[0073] In this embodiment, when the preprocessing result meets the target requirements, the preprocessing result is used as the target processing result. Or, when the preprocessing result does not meet the target requirements, the target processing result is obtained through the target agent. As a result, for simple tasks (such as playing songs, querying the weather, etc.), the target processing result can be obtained in the preprocessing stage, improving the response speed and reducing resource overhead. For complex tasks, the target processing result is obtained through the target agent, such as performing complex schedule planning and multi-device linkage through the target agent, which can more accurately meet the requirements of complex tasks and improve reliability. In addition, by removing the preset type of tools, its execution path can be controlled to be isolated from the agent, avoiding the ambiguity points of the agent's decision-making, and ensuring the stability and reliability of the system operation.
[0074] Figure 4 It is a schematic diagram according to the third embodiment of the present disclosure. This embodiment provides a task processing device based on artificial intelligence. As Figure 4 shown, the device 400 includes: a first processing module 401, a determination module 402, and a second processing module 403.
[0075] The first processing module 401 is configured to perform a single-round reasoning process on the task request to obtain a preprocessing result; the decision module 402 is configured to make a decision on the preprocessing result to obtain a decision result; the decision result is used to indicate whether the preprocessing result meets the target requirements, and the target agent to be called when the preprocessing result does not meet the target requirements; the second processing module 403 is configured to, in response to the decision result indicating that the preprocessing result does not meet the target requirements, use the target agent to perform a standard reasoning process on the task request to obtain the target processing result of the task request.
[0076] In this embodiment, when the preprocessing result does not meet the target requirements, by using the target agent to perform a standard reasoning process on the task request, unnecessary agent calls can be minimized, thereby improving the response speed and reducing resource overhead.
[0077] In some embodiments, the first processing module 401 is further configured to:
[0078] Adopt a small model to perform a thought chain reasoning on the task request to obtain a reasoning result;
[0079] Based on the reasoning result and preset information, determine a target tool;
[0080] Call the target tool to obtain the preprocessing result.
[0081] In this embodiment, the target tool is determined based on the reasoning results and preset information. This method is compatible with existing resources, and combining multiple types of information to determine the target tool can improve accuracy.
[0082] In some embodiments, the decision module 402 is further configured to:
[0083] The information and associated data of the target tool are input into a pre-trained classification model to obtain the decision result output by the classification model.
[0084] In this embodiment, a classification model is used to determine whether the preprocessing results meet the target requirements and the target intelligent agent. This allows for simple and efficient classification, enabling high-precision traffic distribution.
[0085] In some embodiments, the second processing module 403 is further configured to:
[0086] Retrieve the historical tools corresponding to the task request;
[0087] Remove preset type tools from the historical tools and the target tools to obtain filtered tools;
[0088] The filtered tool and the task request are sent to the target agent so that the target agent can perform a standard inference process to obtain the target processing result of the task request.
[0089] In this embodiment, by removing preset types of tools, tools that are difficult for the agent to understand can be filtered out, reducing the agent's processing burden and improving the accuracy of the processing results. By acquiring historical tools, processing them, and then sending them to the target agent, the target agent can refer to historical data, reducing the illusion problem and improving accuracy.
[0090] In some embodiments, the device 400 further includes:
[0091] The third processing module is used to, in response to the decision result indicating that the preprocessing result meets the target requirements, use the preprocessing result as the target processing result of the task request.
[0092] In this embodiment, when the preprocessing result meets the target requirements, the preprocessing result is used as the target processing result without calling the intelligent agent, which can reduce latency and cost.
[0093] Figure 5 This is a schematic diagram based on the fourth embodiment of the present disclosure, which provides a task processing system based on artificial intelligence. Figure 5 As shown, the system 500 includes: a front-end executor 501 and a target intelligent agent 502.
[0094] The pre-processing agent 501 is used to perform a single-round reasoning process on the task request to obtain a pre-processing result; to make a decision on the pre-processing result to obtain a decision result; the decision result is used to indicate whether the pre-processing result meets the target requirements, and the target agent to be invoked when it does not meet the target requirements; the target agent 502 is used to perform a standard reasoning process on the task request to obtain the target processing result of the task request when the decision result indicates that the pre-processing result does not meet the target requirements.
[0095] In this embodiment, when the preprocessing result does not meet the target requirements, the target agent performs a standard reasoning process on the task request, which can minimize unnecessary agent calls, thereby improving response speed and reducing resource consumption.
[0096] It is understood that the same or similar content in different embodiments of this disclosure can be referred to each other.
[0097] It is understood that the terms "first" and "second" in the embodiments of this disclosure are only used for distinction and do not indicate the degree of importance or the order of events.
[0098] It is understandable that, unless otherwise specified, the order of steps in the process indicates that the temporal relationship between these steps is not limited.
[0099] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.
[0100] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.
[0101] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device 600 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.
[0102] like Figure 6As shown, the electronic device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. The RAM 603 may also store various programs and data required for the operation of the electronic device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.
[0103] Multiple components in electronic device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of displays, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows electronic device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0104] The computing unit 601 can be a variety of general-purpose and / or proprietary processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various proprietary artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as AI-based task processing methods. For example, in some embodiments, the AI-based task processing method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the AI-based task processing method described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform AI-based task processing methods by any other suitable means (e.g., by means of firmware).
[0105] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), proprietary integrated circuits (ASICs), proprietary standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a proprietary or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.
[0106] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, a dedicated computer, or other programmable task processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0107] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
[0108] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, 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 sound input, voice input, or tactile input).
[0109] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.
[0110] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.
[0111] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.
[0112] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.
Claims
1. A task processing method based on artificial intelligence, comprising: Perform a single round of reasoning on the task request to obtain the pre-processing results; The results of the preprocessing are used to make a decision to obtain a decision result; The decision result is used to indicate whether the preprocessing result meets the target requirements, and the target agent to be invoked when it does not meet the target requirements; In response to the decision result indicating that the preprocessing result does not meet the target requirements, the target agent performs a standard reasoning process on the task request to obtain the target processing result of the task request.
2. The method according to claim 1, wherein, The process of performing a single round of reasoning on the task request to obtain the pre-processing result includes: A small model is used to perform thought chain reasoning on the task request to obtain the reasoning result; Based on the reasoning results and preset information, the target tool is determined; The target tool is invoked to obtain the preprocessing results.
3. The method according to claim 2, wherein, The step of making a decision based on the preprocessing results to obtain a decision result includes: The information and associated data of the target tool are input into a pre-trained classification model to obtain the decision result output by the classification model.
4. The method according to claim 2, wherein, The step of using the target agent to perform a standard inference process on the task request to obtain the target processing result of the task request includes: Retrieve the historical tools corresponding to the task request; Remove preset type tools from the historical tools and the target tools to obtain filtered tools; The filtered tool and the task request are sent to the target agent so that the target agent can perform a standard inference process to obtain the target processing result of the task request.
5. The method according to claim 2, further comprising: In response to the decision result indicating that the preprocessing result meets the target requirements, the preprocessing result is used as the target processing result of the task request.
6. A task processing device based on artificial intelligence, comprising: The first processing module is used to perform a single round of reasoning on the task request in order to obtain the pre-processing results; The decision module is used to make decisions based on the preprocessing results to obtain decision results; The decision result is used to indicate whether the preprocessing result meets the target requirements, and the target agent to be invoked when it does not meet the target requirements; The second processing module is used to respond to the decision result indicating that the preprocessing result does not meet the target requirements, and to use the target agent to perform a standard reasoning process on the task request to obtain the target processing result of the task request.
7. An artificial intelligence-based task processing system, comprising: The pre-processing unit is used to perform a single round of reasoning on the task request in order to obtain the pre-processing result; The results of the preprocessing are used to make a decision to obtain a decision result; The decision result is used to indicate whether the preprocessing result meets the target requirements, and the target agent to be invoked when it does not meet the target requirements; A target agent is used to perform a standard reasoning process on the task request to obtain the target processing result of the task request when the decision result indicates that the preprocessing result does not meet the target requirements.
8. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5.
9. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-5.
10. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-5.