Information interaction method, device and electronic device based on large language model

Abstract

The present disclosure provides an information interaction method, device and electronic device based on a large language model, in the field of artificial intelligence technology, especially in the field of deep learning, large language models and other technical fields, which can be applied to scenarios such as artificial intelligence content generation. The specific implementation scheme is: in response to obtaining demand description information, using a large language model to process the demand description information, and obtaining a task execution strategy that matches the demand description information, wherein the task execution strategy includes multiple demand task attributes, and at least one of the multiple demand task attributes matches the demand intent represented by the demand description information; according to the demand task attributes in the task execution strategy, determining the strategy execution result related to the task execution strategy; and generating feedback information according to the strategy execution result, and displaying the feedback information.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, in particular to technical fields such as deep learning and large language models, and can be applied to scenarios such as artificial intelligence content generation. Background Art

[0002] With the rapid development of artificial intelligence technology, Internet service platforms can train deep learning models based on actual business needs, and use pre-trained deep learning models to help process business information, thereby improving business service efficiency. Summary of the invention

[0003] The present disclosure provides an information interaction method, device, electronic device and storage medium based on a large language model.

[0004] According to one aspect of the present disclosure, there is provided an information interaction method based on a large language model, comprising: in response to acquiring requirement description information, processing the requirement description information using a large language model to obtain a task execution strategy that matches the requirement description information, wherein the task execution strategy includes multiple requirement task attributes, and at least one of the multiple requirement task attributes matches the requirement intent represented by the requirement description information; determining a policy execution result related to the task execution strategy based on the requirement task attributes in the task execution strategy; and generating feedback information based on the policy execution result, and displaying the feedback information.

[0005] According to another aspect of the present disclosure, a method for information interaction based on a large language model is provided, including: in response to demand description information input by a target object, using at least one intelligent agent among multiple intelligent agents to process the demand description information and obtain feedback information, wherein the intelligent agent is suitable for performing the following operations: in response to obtaining the demand description information, using a large language model to process the demand description information and obtain a task execution strategy that matches the demand description information, wherein the task execution strategy includes multiple demand task attributes, and at least one of the multiple demand task attributes matches the demand intent represented by the demand description information; determining a strategy execution result related to the task execution strategy based on the demand task attributes in the task execution strategy; and generating feedback information based on the strategy execution result, and displaying the feedback information; using at least two of the multiple intelligent agents to implement at least one information interaction based on the feedback information, generating interactive feedback information; and displaying the interactive feedback information.

[0006] According to another aspect of the present disclosure, there is provided an information interaction device based on a large language model, including: a task execution strategy acquisition module, for responding to the acquisition of requirement description information, processing the requirement description information by using a large language model, and obtaining a task execution strategy that matches the requirement description information, wherein the task execution strategy includes multiple requirement task attributes, and at least one of the multiple requirement task attributes matches the requirement intent represented by the requirement description information; a strategy execution result determination module, for determining a strategy execution result related to the task execution strategy according to the requirement task attributes in the task execution strategy; and a feedback module, for generating feedback information according to the strategy execution result, and displaying the feedback information.

[0007] According to another aspect of the present disclosure, there is provided an information interaction device based on a large language model, including: a feedback information acquisition module, for responding to demand description information input by a target object, using at least one of a plurality of intelligent agents to process the demand description information, and obtaining feedback information, wherein the intelligent agent is suitable for performing the following operations: in response to obtaining the demand description information, using a large language model to process the demand description information, and obtaining a task execution strategy that matches the demand description information, wherein the task execution strategy includes a plurality of demand task attributes, and at least one of the plurality of demand task attributes matches the demand intent represented by the demand description information; determining a strategy execution result related to the task execution strategy according to the demand task attributes in the task execution strategy; and generating feedback information according to the strategy execution result, and displaying the feedback information; an interactive feedback information generation module, for generating interactive feedback information by using at least two of the plurality of intelligent agents to implement at least one information interaction based on the feedback information; and an interactive feedback information display module, for displaying the interactive feedback information.

[0008] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method provided according to an embodiment of the present disclosure.

[0009] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method provided according to an embodiment of the present disclosure.

[0010] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the method provided according to the embodiment of the present disclosure is implemented.

[0011] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0013] Figure 1 An exemplary system architecture to which a method and device for information interaction based on a large language model can be applied according to an embodiment of the present disclosure is schematically shown;

[0014] Figure 2 A flowchart of an information interaction method based on a large language model according to an embodiment of the present disclosure is schematically shown;

[0015] Figure 3A A schematic diagram schematically shows a task execution strategy according to an embodiment of the present disclosure;

[0016] Figure 3B A schematic diagram schematically shows a task execution strategy according to another embodiment of the present disclosure;

[0017] Figure 4A A schematic diagram schematically shows a task execution strategy according to another embodiment of the present disclosure;

[0018] Figure 4B A schematic diagram schematically shows a task execution strategy according to another embodiment of the present disclosure;

[0019] Figure 5 The application scenario diagram of the information interaction method based on the large language model according to the embodiment of the present disclosure is schematically shown;

[0020] Figure 6 The functional framework diagram of the advertising marketing intelligent system suitable for executing the information interaction method based on the large language model according to the embodiment of the present disclosure is schematically shown;

[0021] Figure 7 A framework diagram for executing the information interaction method based on a large language model according to an embodiment of the present disclosure is schematically shown;

[0022] Figure 8 Schematically shows an architecture diagram of an advertising and marketing system suitable for executing an embodiment of the present disclosure;

[0023] Fig. 9 Schematically shows an architecture diagram suitable for executing an intelligent agent according to an embodiment of the present disclosure;

[0024] Fig.10Schematically shows a principle diagram suitable for executing the evolution of an intelligent agent according to an embodiment of the present disclosure;

[0025] Fig.11 Schematically shows a schematic diagram suitable for executing the agent evolution according to another embodiment of the present disclosure

[0026] Fig.12 Schematically shows a flow chart of an information interaction method based on a large language model according to another embodiment of the present disclosure;

[0027] Fig.13 Schematically shows an application scenario diagram of an information interaction method based on a large language model according to another embodiment of the present disclosure;

[0028] Fig.14 A block diagram of an information interaction device based on a large language model according to an embodiment of the present disclosure is schematically shown;

[0029] Fig.15 A block diagram schematically shows an information interaction device based on a large language model according to another embodiment of the present disclosure; and

[0030] Fig.16 A block diagram of an electronic device suitable for implementing an information interaction method based on a large language model according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0031] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0032] In the technical solution disclosed in the present invention, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations, necessary confidentiality measures are taken, and do not violate public order and good morals.

[0033] The embodiments of the present disclosure provide an information interaction method, device, electronic device and storage medium based on a large language model. The information interaction method based on a large language model includes: in response to obtaining demand description information, using a large language model to process the demand description information, and obtaining a task execution strategy that matches the demand description information, wherein the task execution strategy includes multiple demand task attributes, and at least one of the multiple demand task attributes matches the demand intent represented by the demand description information; according to the demand task attributes in the task execution strategy, determining a strategy execution result related to the task execution strategy; and generating feedback information according to the strategy execution result, and displaying the feedback information.

[0034] According to the embodiments of the present disclosure, by utilizing a large language model to process the requirement description information, a task execution strategy is obtained so that some of the requirement task attributes in the task execution strategy match the requirement intention represented by the requirement description information. This allows the task execution strategy containing multiple requirement task attributes to further meet the potential requirement intention not expressed by the requirement description information on the basis of meeting the requirement intention represented by the requirement description information. This allows the strategy execution result to be determined based on the requirement task attributes in the task execution strategy, and further feedback information to be generated. This allows the feedback information to more fully meet the potential requirement intention of the target object, reduce the target object's input operation process of inputting a descriptive text representing the potential requirement intention during the next information interaction process, and improve the efficiency of the target object in obtaining information.

[0035] Figure 1 An exemplary system architecture to which a large language model-based information interaction method and apparatus can be applied according to an embodiment of the present disclosure is schematically shown.

[0036] It should be noted that Figure 1 The examples shown are only examples of system architectures to which the embodiments of the present disclosure can be applied, in order to help those skilled in the art understand the technical content of the present disclosure, but do not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios. For example, in another embodiment, an exemplary system architecture to which the information interaction method and apparatus based on a large language model can be applied may include a terminal device, but the terminal device may implement the information interaction method and apparatus based on a large language model provided by the embodiments of the present disclosure without interacting with a server.

[0037] like Figure 1 As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0038] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients and / or social platform software, etc. (only as examples).

[0039] The terminal devices 101 , 102 , and 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0040] The server 105 may be a server that provides various services, such as a background management server (only an example) that provides support for the content browsed by the user using the terminal devices 101, 102, and 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0041] Server 105 may also be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0042] It should be noted that the information interaction method based on the large language model provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the information interaction device based on the large language model provided in the embodiment of the present disclosure can generally be set in the server 105. The information interaction method based on the large language model provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105. Accordingly, the information interaction device based on the large language model provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105.

[0043] Alternatively, the information interaction method based on the large language model provided in the embodiment of the present disclosure may also be executed by the terminal device 101, 102 or 103. Accordingly, the information interaction apparatus based on the large language model provided in the embodiment of the present disclosure may also be set in the terminal device 101, 102 or 103.

[0044] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0045] Figure 2 The flowchart of the information interaction method based on the large language model according to the embodiment of the present disclosure is schematically shown.

[0046] like Figure 2 As shown, the information interaction method based on the large language model includes operations S210 to S230.

[0047] In operation S210, in response to obtaining the requirement description information, the requirement description information is processed using a large language model to obtain a task execution strategy that matches the requirement description information, wherein the task execution strategy includes multiple requirement task attributes, and at least one of the multiple requirement task attributes matches the requirement intent represented by the requirement description information.

[0048] In operation S220, a policy execution result related to the task execution policy is determined according to the required task attributes in the task execution policy.

[0049] In operation S230, feedback information is generated according to the policy execution result, and the feedback information is displayed.

[0050] According to an embodiment of the present disclosure, the demand description information may be natural language information used to characterize the target object's information acquisition needs. For example, the demand description information may be "how is the recent performance of the xx marketing plan", and the demand description information may characterize the target object's acquisition demand intention for the revenue information of the xx marketing plan.

[0051] According to the embodiments of the present disclosure, the demand description information can be obtained based on the input operation of the target object on the information input box in the interactive interface. However, it is not limited to this, and the demand description information can also be obtained by other means, for example, the sound information expressing the demand description information of the target object can be collected based on the sound collection device, and the demand description information represented by the sound information can be obtained by the speech recognition algorithm. The embodiments of the present disclosure do not limit the specific method of obtaining the demand description information.

[0052] According to an embodiment of the present disclosure, a large language model (LLM) may include a deep learning model trained using a large amount of text data. For example, a large language model may include a model constructed based on a neural network model such as a Transformer model to understand the meaning of a language text and generate a natural language text. A large language model can process natural language tasks. Since a large language model typically contains billions of parameters, large-scale parameters can help the large language model learn complex patterns in natural language data, thereby performing better in natural language processing (NLP) tasks.

[0053] According to an embodiment of the present disclosure, a task execution strategy may represent one or more tasks that need to be performed in order to obtain demand-related information, and the demand-related information may match the demand intent represented by the demand description information, and may also match the potential demand intent represented by the demand description information. Among the multiple demand task attributes of the task execution strategy, a part of the demand task attributes match the demand intent represented by the demand description information, and another part of the demand task attributes match the potential demand intent represented by the demand description information. For example, the multiple demand task attributes may include demand task attribute A, demand task attribute B, and demand task attribute C, and the demand task attribute A matches the demand intent represented by the demand description information, and the demand task attribute B and the demand task attribute C match the potential demand intent represented by the demand description information. The number of demand task attributes in the task execution strategy may be greater than the number of demand intent represented by the demand description information. It should be noted that the potential demand intent may include a deeper level of demand intent relative to the demand intent based on the demand intent represented by the demand description information.

[0054] According to the embodiments of the present disclosure, the large language model can be pre-trained based on the business application data, so that the large language model can predict the demand task attributes that match the demand intent and potential demand intent (or potential demand intent) based on the demand description text expressed in natural language. Alternatively, the corresponding prompt information can be obtained or edited based on the demand description information, and the prompt information is processed by the large language model to generate demand task attributes that match the potential demand intent.

[0055] According to an embodiment of the present disclosure, the requirement task attributes may include attribute information such as task configuration parameters and task type for obtaining requirement-related information that matches the requirement intention and potential requirement intention. The requirement task can be generated based on the requirement task attributes, and part or all of the requirement-related information can be obtained by executing the requirement task.

[0056] For example, the requirements task attributes may include attribute parameters such as the table name of the data table that needs to be obtained, the data table generation time, the data name, etc., so that the requirements task can be generated according to the requirements task attributes. The requirements task may include a query statement for querying data. By executing the query statement, the data in the data table that matches the requirements task attributes can be obtained, and the requirements task execution result can be obtained.

[0057] According to the embodiments of the present disclosure, according to the required task attributes in the task execution strategy, the strategy execution result related to the task execution strategy is determined, the required task can be generated by the required task attributes, and the task execution result can be obtained by executing the required task. The task execution result can include information matching the required intention and information matching the potential required intention. Thus, the strategy execution result can be obtained according to multiple task execution results.

[0058] According to an embodiment of the present disclosure, generating feedback information according to the policy execution result may include using the policy execution result as feedback information. However, it is not limited thereto, or it may also include filling a preset configuration template based on the policy execution result to obtain feedback information. The embodiment of the present disclosure does not limit the specific method of determining the feedback information, as long as it can characterize the policy execution result.

[0059] According to an embodiment of the present disclosure, using a large language model to process requirement description information to obtain a task execution strategy that matches the requirement description information may include: using a large language model to process the requirement description information to obtain multi-level requirement intentions; and determining a task execution strategy based on the multi-level requirement intentions.

[0060] According to the embodiments of the present disclosure, a multi-level demand intent may include multiple demand intents that are continuously expanded in the demand hierarchy, and the demand intent of the current level may be the intent information after the intention of the demand intent of the previous level is expanded. The large language model can be pre-trained by selecting multiple rounds of interaction information as samples, so that the pre-trained large language model can predict the multi-level demand intent that matches the demand description information.

[0061] According to an embodiment of the present disclosure, determining a task execution strategy based on multi-level demand intentions may include determining demand task attributes corresponding to demand intentions at each level in the multi-level demand intentions, and then obtaining the task execution strategy based on multiple demand task attributes.

[0062] According to the embodiments of the present disclosure, multi-level demand intention recognition is performed on demand description information through a large language model, and expanded deeper-level demands can be mined from the demand description information to reduce the operational steps for the target object to input deep potential demands through multiple rounds of interaction.

[0063] According to an embodiment of the present disclosure, using a large language model to process requirement description information to obtain multi-level requirement intentions may include: for the current level, using a large language model to process the requirement description information and the requirement intention of the previous level to obtain the requirement intention of the current level, wherein the requirement intention of the first level in the multi-level requirement intention is obtained by processing the requirement description information using a large language model.

[0064] According to the embodiments of the present disclosure, based on the large language model's strong natural language semantic understanding and text prediction capabilities, the large language model can be used iteratively to process the demand intention of the previous level that has been detected, and then predict the demand intention of the current level. This can achieve continuous expansion and mining of demand intentions, and the obtained multi-level demand intentions contain potential demand intentions represented by demand description information.

[0065] According to an embodiment of the present disclosure, using a large language model to process demand description information to obtain multi-level demand intent may also include: using a large language model to process demand description information, and obtaining multi-level demand intent based on associated query information queried based on the demand description information.

[0066] According to the embodiments of the present disclosure, part or all of the demand description information can be used as query keywords (or queries) to search for related query information using search tools or query statements. The related query information can include one or more of the context information and professional knowledge related to the demand description information, thereby helping the large language model to understand the semantics expressed by the demand description information and predict multi-level demand intentions including potential demand intentions.

[0067] According to the embodiments of the present disclosure, a knowledge base related to business services can be constructed so that related query information can be queried from the knowledge base according to the demand description information, thereby improving the relevance of the related query information and the demand description information.

[0068] According to an embodiment of the present disclosure, before determining the task execution strategy according to the multi-level demand intention, the information interaction method may further include: displaying at least one demand intention among the multi-level demand intention; and in response to a selection operation for at least one demand intention, determining the demand intention associated with the selection operation;

[0069] According to an embodiment of the present disclosure, determining the task execution strategy according to the multi-level demand intentions may include: determining the task execution strategy according to the demand intention associated with the selection operation in the multi-level demand intentions.

[0070] According to the embodiments of the present disclosure, the matching degree between the multi-level demand intentions and the real intentions of the target objects can be improved by displaying the demand intentions for confirmation by the target objects, so that the obtained feedback information can truly meet the needs of the target objects.

[0071] According to an embodiment of the present disclosure, determining a policy execution result related to a task execution strategy based on the requirement task attributes in the task execution strategy may include: determining a completed requirement task attribute that has been processed from multiple requirement task attributes of the current task execution strategy; using a large language model to process the completed requirement task attribute and the task execution results associated with the completed requirement task attribute to obtain the requirement task attribute to be processed; updating the current task execution strategy based on the requirement task attribute to be processed to obtain a new task execution strategy; and determining the policy execution result based on the new task execution strategy.

[0072] According to an embodiment of the present disclosure, the current task execution strategy may include an updated task execution strategy or an unupdated task execution strategy. Completing the required task attribute may mean that a required task has been generated according to the completed task attribute and the required task has been executed.

[0073] According to an embodiment of the present disclosure, using a large language model to process completion requirement task attributes can include inputting some or all of the completion requirement task attributes that have been executed in the current task execution strategy, as well as task execution results corresponding to some or all of the completion requirement task attributes, into the large language model, so that the large language model can understand the execution status of the current task execution strategy, and then determine whether the current task execution strategy needs to be updated, so that the new task execution strategy can further match the requirement intention and potential requirement intention represented by the requirement description information.

[0074] According to the embodiments of the present disclosure, a new task execution strategy can be used as the current task execution strategy, and the current task execution strategy can be updated in an iterative optimization manner through the current completion requirement task attributes and the task execution results corresponding to the current completion requirement task attributes, so that the strategy execution results corresponding to the most recently updated task execution strategy can meet the multi-level demand intentions of the target object, save the interaction time required for the target object to obtain information, and at the same time save the communication space and computing resources generated by the information interaction.

[0075] According to an embodiment of the present disclosure, updating the current task execution strategy according to the attributes of the pending demand tasks may include: updating the attributes of unfinished demand tasks that have not been processed in the current task execution strategy according to the attributes of the pending demand tasks.

[0076] According to the embodiments of the present disclosure, the attributes of unfinished demand tasks that have not yet been processed in the current task execution strategy can be updated through the attributes of the demand tasks to be processed, so that the attributes of the unfinished demand tasks in the new task execution strategy obtained after the update can be adapted to the potential demand intentions of the target object, and thus the task execution results can be obtained according to the updated attributes of the demand tasks to be processed. In this way, the self-reflection of the task execution strategy can be realized according to the large language model, and then the attributes of the unfinished demand tasks that have not yet been processed can be corrected in a timely manner, so as to timely adjust the current task execution strategy, avoid the waste of computing resources caused by the execution of demand task attributes that differ greatly from the demand intention, and avoid generating feedback information based on irrelevant task execution results that affect the user experience of the target object.

[0077] According to an embodiment of the present disclosure, updating the current task execution strategy according to the attributes of the pending demand tasks may also include: updating the completed demand task attributes in the current task execution strategy according to the attributes of the pending demand tasks.

[0078] According to the embodiments of the present disclosure, by updating the completion requirement task attributes in the current task execution strategy, it can be detected based on the language understanding ability of the large language model that the completion requirement attribute task that has been processed and completed has little correlation with the demand intention or potential demand intention, so that the task attributes to be processed can be output to replace the completion requirement task attributes, thereby achieving self-reflection to correct the predicted demand judgment errors, or timely adjust the current task execution strategy according to the task execution results associated with the completion task attributes, so that the new task execution strategy can be more adapted to the demand intention and potential demand intention of the target object, and avoid generating feedback information based on irrelevant task execution results that affect the target object's usage experience.

[0079] In one embodiment of the present disclosure, some or all completed demand task attributes in the current task execution strategy may be updated according to the pending demand tasks, and some or all uncompleted demand task attributes in the current task execution strategy may be updated.

[0080] According to an embodiment of the present disclosure, the current task execution strategy may include the execution dependencies between the current demand task attributes. Updating the current task execution strategy based on the demand task attributes to be processed may also include updating the execution dependencies contained in the current task execution strategy, so that the new task execution strategy can execute the demand task according to the new execution dependencies, so as to improve the adaptability of the final strategy execution result to the demand intention and potential demand intention of the target object.

[0081] According to an embodiment of the present disclosure, a task execution strategy may be characterized based on a task execution topology, nodes in the task execution topology may characterize required task attributes, and edge relationships between required task attributes may characterize execution dependencies.

[0082] Figure 3A A schematic diagram schematically shows a task execution strategy according to an embodiment of the present disclosure.

[0083] Figure 3B A schematic diagram schematically shows a task execution strategy according to another embodiment of the present disclosure.

[0084] like Figure 3A As shown, the current task execution strategy 300A may include multiple demand task attributes, and the multiple demand task attributes are respectively A demand task attribute 311, B demand task attribute 321, C demand task attribute 322 and D demand task attribute 331. The task execution strategy 300A can be represented by a task execution topology, and the A demand task attribute 311, B demand task attribute 321, C demand task attribute 322 and D demand task attribute 331 are respectively nodes in the task execution topology, and the execution dependency relationship between the A demand task attribute 311, B demand task attribute 321, C demand task attribute 322 and D demand task attribute 331 is represented by an edge relationship. The task execution strategy 300A can represent the logical relationship of the demand task attributes being processed through the task execution topology. Among them, the A demand task attribute 311 and the B demand task attribute 321 can be completed demand task attributes that have been processed, and the C demand task attribute 322 and the D demand task attribute 331 can be unfinished demand task attributes that have not yet been processed.

[0085] In one embodiment of the present disclosure, Figure 3A and Figure 3B As shown, the large language model can be used to process the A requirement task attribute 311 and the B requirement task attribute 321, as well as the task execution results associated with the A requirement task attribute 311 and the B requirement task attribute 321, to obtain the task attribute to be processed and the update indication information of the task attribute to be processed. The update indication information of the task attribute to be processed can indicate to update the B requirement task attribute 321 in the current task execution strategy 300A. Therefore, according to updating the current task execution strategy 300A with the task attribute to be processed, a new task execution strategy 300B can be obtained. The new task execution strategy 300B can include the B' requirement task attribute 321'.

[0086] Figure 4A A schematic diagram schematically shows a task execution strategy according to yet another embodiment of the present disclosure.

[0087] Figure 4BA schematic diagram schematically shows a task execution strategy according to yet another embodiment of the present disclosure.

[0088] like Figure 4A As shown, the current task execution strategy 400A may include multiple demand task attributes, and the multiple demand task attributes are respectively A1 demand task attribute 411, B1 demand task attribute 421, C1 demand task attribute 422 and D1 demand task attribute 431. The task execution strategy 400A can be represented by a task execution topology, and the A1 demand task attribute 411, B1 demand task attribute 421, C1 demand task attribute 422 and D1 demand task attribute 431 are nodes in the task execution topology, and the execution dependency relationship between the A1 demand task attribute 411, B1 demand task attribute 421, C1 demand task attribute 422 and D1 demand task attribute 431 is represented by an edge relationship. The task execution strategy 400A can represent the logical relationship of the demand task attributes being processed through the task execution topology. Among them, the A1 demand task attribute 411, B1 demand task attribute 421 and C1 demand task attribute 422 can be completed demand task attributes that have been processed, and the D1 demand task attribute 431 can be an unfinished demand task attribute that has not yet been processed.

[0089] In one embodiment of the present disclosure, Figure 4A and Figure 4B As shown, the large language model can be used to process the A1 requirement task attribute 411, the B1 requirement task attribute 421 and the C1 requirement task attribute 422, as well as the task execution results associated with the A1 requirement task attribute 411, the B1 requirement task attribute 421 and the C1 requirement task attribute 422, to obtain the pending task attribute and the update indication information of the pending task attribute. The update indication information of the pending task attribute can indicate the update of the B1 requirement task attribute 421, the C1 requirement task attribute 422 and the D1 requirement task attribute 431 in the current task execution strategy 400A, and the update indication information of the pending task attribute can also indicate the update of the execution dependency in the current task execution strategy 400A. Therefore, according to the pending task attribute, the current task execution strategy 400A is updated, and a new task execution strategy 400B can be obtained. The new task execution strategy 400B may include A1 requirement task attribute 411, B1' requirement task attribute 421', C1' requirement task attribute 422' and D1' requirement task attribute 431', and the execution dependency relationship between the requirement task attributes of the new task execution strategy 400B is also updated according to the update prompt information output by the large language model. The obtained new task execution strategy 400B can be combined with the task execution results of the completed requirement task attributes for timely reflection and adjustment, so that the new task execution strategy 400B can meet the requirement intention of the target object.

[0090] According to an embodiment of the present disclosure, a new task execution strategy includes new execution dependencies between required task attributes.

[0091] According to an embodiment of the present disclosure, according to a new task execution strategy, determining the strategy execution result may include: determining the first target requirement task according to the first target requirement task attribute, and the first dependent task execution result associated with the first dependent requirement task attribute, wherein the new task execution strategy includes the first target requirement task attribute and the first dependent requirement task attribute, and there is an execution dependency relationship between the first target requirement task attribute and the first dependent requirement task attribute; calling the first target service resource associated with the first target requirement task attribute to execute the first target requirement task, and obtaining the task execution result associated with the first target requirement task attribute, and the strategy execution result includes the task execution result associated with the first target requirement task attribute.

[0092] According to an embodiment of the present disclosure, the first target requirement task attribute may include a requirement task attribute that needs to be processed, and the first dependent requirement task attribute may include a requirement task attribute of a previous level associated with the first target requirement task attribute, or may also include a requirement task attribute associated with the first target requirement task attribute at an interval of one level or multiple levels. Those skilled in the art may determine the first dependent requirement task attribute based on actual needs, as long as it satisfies a direct or indirect execution dependency relationship with the first target requirement task attribute.

[0093] According to an embodiment of the present disclosure, determining the first target requirement task based on the first target requirement task attributes and the first dependent task execution result associated with the first dependent requirement task attributes can include generating the first target requirement task based on the task matching parameters included in the first target requirement task attributes and the first dependent task execution result that has been generated.

[0094] According to an embodiment of the present disclosure, the first target requirement task attribute may include a target service resource identifier, and the target service resource identifier is suitable for characterizing the association relationship between the first target requirement task attribute and the first target service resource. By calling the first target service resource to execute the first target requirement task through the target service resource identifier, it is possible to avoid the target object from executing the first target requirement task by sending a resource call instruction or performing a resource selection operation, thereby realizing automatic calling of service resources and saving the target object's interactive operation time.

[0095] According to an embodiment of the present disclosure, the first target requirement task attribute is characterized based on a prompt template. According to the first target requirement task attribute and the first dependent task execution result associated with the first dependent requirement task attribute, determining the first target requirement task includes: updating the first target requirement task attribute according to the first dependent task execution result associated with the first dependent requirement task attribute to obtain the first target task prompt information; and using the large language model to process the first target task prompt information to obtain the first target requirement task.

[0096] According to an embodiment of the present disclosure, a prompt template (or prompt template) may be a task configuration parameter used to help a large language model understand the attributes of a first target requirement task. The prompt template may be based on a prompt tag sequence that controls the large language model for accurate prediction. The prompt tag sequence may include any type of prompt tags such as characters, fields, and text. The first target task attributes are filled in according to the execution result of the first dependent task, and the first target task prompt information may be obtained. Thus, by processing the first target task prompt information using a large language model, the obtained first target requirement task may more accurately match the demand intent and potential demand intent. In this way, after executing the first target requirement task, the task execution result associated with the first target requirement task attributes obtained may more accurately match the needs of the target object.

[0097] According to an embodiment of the present disclosure, the task execution strategy also includes an execution dependency relationship between multiple demand task attributes. According to the demand task attributes in the task execution strategy, determining the policy execution result related to the task execution strategy may include: generating a second target demand task according to the second target demand task attribute and the task execution result associated with the second dependent demand task attribute, wherein the task execution strategy includes the second dependent demand task attribute and the second target demand task attribute, and there is an execution dependency relationship between the second dependent demand task attribute and the second target demand task attribute; and calling the second target service resource associated with the second target demand task attribute to execute the second target demand task, and obtaining the task execution result associated with the second target demand task attribute, wherein the policy execution result includes the task execution result associated with the second target demand task attribute.

[0098] According to the embodiments of the present disclosure, without updating the task execution policy, the second target task can be generated in sequence from the second target demand task attribute in the task execution policy, according to the execution dependency relationship between the multiple demand task attributes indicated by the task execution policy, and according to the second dependent demand task attribute having an execution dependency relationship with the second target demand task attribute, and the task execution result associated with the second dependent demand task attribute, so as to realize the automatic generation of the target task to be executed, and avoid filling in the task configuration parameters to generate the operation steps of the task to be executed. And by calling the second target service resource associated with the second target demand task attribute to execute the second target demand task, it is possible to realize accurate and automatic calling of resources, and avoid the problem of long resource calling operation time caused by manual operation selection.

[0099] According to an embodiment of the present disclosure, the second target requirement task attribute includes a target service resource identifier, and the target service resource identifier is suitable for characterizing the association relationship between the second target requirement task attribute and the second target service resource.

[0100] According to an embodiment of the present disclosure, the second target requirement task attribute is characterized based on a prompt template; wherein, according to the second target requirement task attribute and the task execution result associated with the second dependent requirement task attribute, generating the second target requirement task includes: updating the second target requirement task attribute according to the task execution result associated with the second dependent requirement task attribute to obtain second target task prompt information; and using a large language model to process the second target task prompt information to obtain the second target requirement task.

[0101] According to an embodiment of the present disclosure, a prompt template (or prompt template) may be a task configuration parameter used to help a large language model understand the attributes of a second target requirement task. The prompt template may be based on a prompt tag sequence that controls the large language model for accurate prediction. The prompt tag sequence may include any type of prompt tags such as characters, fields, and text. The second target task attributes are filled in according to the execution result of the second dependent task, and the second target task prompt information may be obtained. Thus, by processing the second target task prompt information with a large language model, the obtained second target requirement task may more accurately match the requirement intent and potential requirement intent. In this way, after executing the second target requirement task, the task execution result associated with the second target requirement task attribute may more accurately match the requirements of the target object.

[0102] According to an embodiment of the present disclosure, the information interaction method based on the large language model may further include: displaying the target task execution result, wherein the target task execution result includes the task execution result associated with the first target requirement task attribute.

[0103] According to an embodiment of the present disclosure, by displaying the task execution results associated with the first target demand task attribute, the target object can browse the currently obtained demand-related information in a timely manner, thereby improving the timeliness of information display.

[0104] According to the embodiments of the present disclosure, the target object can also perform any type of operations such as selection operations and editing operations on the displayed target task execution results, so that the target object can promptly intervene in the execution process of the current task execution strategy, and then adaptively adjust the task execution strategy according to the current needs of the target object, so that the target object can obtain information that matches its needs.

[0105] According to an embodiment of the present disclosure, the target task execution result may further include a task execution result associated with the second target requirement task attribute.

[0106] According to the embodiments of the present disclosure, the target object can browse the displayed task execution results associated with the second target requirement task attributes in a timely manner, and can also perform any type of operations such as selection operations and editing operations on the task execution results associated with the second target requirement task attributes, so that the target object can promptly intervene in the execution process of the current task execution strategy, and then adaptively adjust the task execution strategy according to the current needs of the target object, so that the target object can obtain information that matches its needs.

[0107] Figure 5 The application scenario diagram of the information interaction method based on the large language model according to the embodiment of the present disclosure is schematically shown.

[0108] like Figure 5 As shown, in the application scenario 500, the target object can input the demand description information 501 into the large language model M510 based on the input demand description information 501 "how is the overall performance of the account", and output the task execution strategy 510. According to the first target demand task attribute 511 "use the report to query the broadcast, and obtain the basic delivery data for the marketing plan 1111xxxxx" in the task execution strategy 510, the first target demand task, such as the data collection task 521, can be generated. By calling the target service resource associated with the first target demand task attribute 511, such as the data collection plug-in set 531 to execute the data collection task 521, the first target task execution result can be obtained (for example, it can be "I have obtained the basic delivery data XX").

[0109] like Figure 5As shown, according to the second target requirement task attribute 512 "Use report analysis broadcast, analyze report data, understand current advertising effect, and conduct comparative analysis" in the task execution strategy 510, and the first target task execution result, a second target requirement task, such as data analysis task 522, can be generated. By calling the target service resource associated with the second target requirement task attribute 512, such as the data analysis plug-in set 532 to execute the data analysis task 522, the second target task execution result (for example, "I have completed the analysis of reimbursement data XX") can be obtained.

[0110] like Figure 5 As shown, according to the third target requirement task attribute 513 "Use the effect analysis plug-in to perform delivery effect diagnosis analysis" in the task execution strategy 510 and the second target task execution result, a third target requirement task, such as effect diagnosis analysis task 523, can be generated. By calling the target service resource associated with the third target requirement task attribute 513, such as the effect diagnosis analysis plug-in set 533 to execute the data analysis task 523, the third target task execution result can be obtained (for example, it can be "I have completed the effect diagnosis and found that there is a problem with CTR (Click-Through-Rate)").

[0111] like Figure 5 As shown, according to the fourth target requirement task attribute 514 "using optimized broadcasts to analyze how to optimize accounts" in the task execution strategy 510 and the third target task execution result, a fourth target requirement task, such as a generative rewriting and recommendation task 524, can be generated. By calling the target service resource associated with the fourth target requirement task attribute 514, such as the generative optimization plug-in set 534 to execute the generative rewriting and recommendation task 524, the fourth target task execution result can be obtained (for example, "using optimized broadcasts to analyze how to improve click-through rate can be xxx").

[0112] like Figure 5 As shown, according to the fifth target requirement task attribute 515 "Summarize the analysis results and end the task using the end command" in the task execution strategy 510, and the first target task execution results to the fourth target task execution results, the fifth target requirement task can be generated, and the fifth target requirement task can be executed based on the large language model M510 as the fifth target service resource. The first target task execution results to the fourth target task execution results can be summarized and analyzed to obtain the fifth target task execution result, and the fifth target task execution result can be used as the strategy execution result.

[0113] According to the embodiments of the present disclosure, the task execution results can be characterized based on natural language descriptions, so that during the execution of the task execution strategy, the execution of multiple demand tasks can be completed in an internal dialogue interaction manner, which makes it easier to call the understanding and generative capabilities of the large language model and improves the readability of the task execution results. Therefore, the target object only needs to issue an open-ended expression demand description of the query information "How is the overall performance of the account", and the end-to-end one-stop data analysis process of basic reports, report analysis, problem diagnosis, and action optimization can be successfully completed, thereby improving the quality of business services.

[0114] According to the embodiments of the present disclosure, service resources can realize the integration of multiple types of plug-in resources, facilitate real-time calling of multiple types of plug-ins through required task attributes, and can adjust the task execution strategy through continuous iterative optimization of the task execution strategy until the demand intention of the target object is met.

[0115] The information interaction method provided according to the embodiment of the present disclosure can be implemented in the manner shown in Table 1. For example, the information interaction method provided according to the embodiment of the present disclosure can be implemented through an online solution or an offline solution.

[0116] Table 1

[0117]

[0118] According to the information interaction method based on the large language model provided by the embodiment of the present disclosure, an advertising marketing intelligent system of "perception + thinking + action" can be constructed. The advertising marketing intelligent system includes a reasoning planning subsystem, a memory subsystem, a perception subsystem, an action subsystem, etc. The advertising marketing intelligent system can realize the global capability coordination of the advertising marketing business with the large language model as the core. The functions of each subsystem are described as follows.

[0119] Planning subsystem: Based on the large language model, deep reasoning and thinking are triggered through the task execution strategy paradigm, and reinforcement learning is performed based on the marketing knowledge base and prompt information to achieve task planning for given marketing business goals. It is specifically responsible for task planning, subtask decomposition, reflection and correction of the execution process, and goal-oriented collaborative task achievement. The planning subsystem can have multiple switchable thinking paradigms. For example, the task execution strategy can be continuously optimized iteratively through method (a) for multiple rounds of task execution strategy planning, task execution, task analysis, reflection on task execution results, and task execution strategy update. Among them, the reflection process can process the task execution results output by the previous round based on the large language model to determine whether the task nodes of the remaining rounds can meet the needs of users, and then the process nodes of the remaining rounds can be adjusted in time, such as adding, modifying, and deleting task nodes, so as to achieve further task planning updates. Method (b) can execute task execution strategies for multiple rounds of subtask decomposition, task priority sorting, and high-priority task execution paradigms. Method (c) can execute topology planning for tasks of a one-time global task execution strategy, and execute tasks in topological order. Among them, methods (b) and (c) can improve the efficiency of responding to user input information without going through the reflection process.

[0120] Memory subsystem: stores information perceived from the environment and uses recorded data to promote future response accuracy actions. The memory module in the memory subsystem can help the large language model accumulate experience, self-evolve, and complete tasks in a more consistent, reasonable and effective way. The memory subsystem contains a short-term memory bank and a working memory bank. The working memory bank is used to store the contextual information of the task (such as the intermediate reasoning process and results), which can be stored and retrieved based on a vector database. The long-term memory bank can store a large amount of marketing business domain knowledge. For example, it can include personalized information picked up by customers, customer behavior habits, etc., pre-built business knowledge graphs, historical analysis knowledge, etc., as well as authoritative theoretical and case knowledge in various fields.

[0121] Perception subsystem: The perception space can be expanded from pure text to multimodal fields including text, vision and hearing, so that it can more effectively obtain and utilize more information from the network environment. For example, advertising copy, advertising pictures, advertising videos, rich media information of customer landing pages, etc. can all be perceived by the intelligent body of the advertising marketing intelligent system to provide valuable business information. It can access and call relevant landing pages based on the user's input information, and collect information corresponding to the needs accordingly, or it can also perform action tasks such as advertising delivery.

[0122] Action subsystem: The core of the action subsystem is the use of tools. Tools can expand the action space of the intelligent agent built on the large language model. With the help of tools, the intelligent agent built on the large language model can use various external resources in the reasoning and planning stages. The tools used by the action subsystem include but are not limited to the following tools (also called plug-ins, service resources):

[0123] Text tools: Based on external natural language models or secondary encapsulated LLM tools, improve the quality of text in terms of fluency, relevance, diversity, professionalism and controllability.

[0124] Knowledge retrieval tools: Search-based tools can increase the scope and quality of knowledge accessible to agents with the help of external databases, knowledge graphs, and web pages.

[0125] Business domain tools: Tools in specific fields can enhance the agent’s expertise in the corresponding field and perform specific business actions.

[0126] Natural language to query tool: LLM-based controller that can generate query statements to query the database, or convert user queries into search requests and use search engines to obtain the desired results. Interpret the code, or generate corresponding query statements based on input information, and execute query statements for data retrieval.

[0127] API (interface) tools: APIs of various capabilities, after given structured description information, can be regarded as available tools for Agents.

[0128] Code interpreter: Use the Python interpreter and LaTeX compiler to improve its performance in complex tasks and mathematical calculation tasks.

[0129] Multimodal tools: Use a variety of AI models, such as discriminative models or generative models for images / video / audio / multimodal, to comprehensively complete complex tasks.

[0130] Self-built tools: LLM-based agents can create tools by generating executable programs or integrating existing tools into more powerful tools, and can learn to self-debug and continuously build their own tool library.

[0131] Feedback tool: Feedback information is also abstracted into a tool. Agents can actively or passively seek feedback information intervention to better meet task goals, advance task progress, and deliver high-quality results. It can be displayed on the front end based on the interaction of the background, and an access interface can be provided for users or maintenance personnel to achieve the advancement of tasks.

[0132] Perception-action tools: Considered a key paradigm, it combines model intelligence with the physical world. Use tools that actively perceive, understand, and interact with the physical environment, such as observation, manipulation, and navigation. It can connect to some devices and equipment in the physical world and control devices and equipment based on feedback information. By using various tool plug-in capabilities, the data, knowledge, and capabilities of the advertising and marketing business domain can be organically integrated to achieve autonomous decision-making, select appropriate plug-in combinations, and complete marketing business goals and tasks.

[0133] Figure 6 The functional framework diagram of an advertising marketing intelligent system suitable for executing the information interaction method based on a large language model according to an embodiment of the present disclosure is schematically shown.

[0134] like Figure 6 As shown, the intelligent agent 600 can provide feedback information on the open description of the target object in natural language during each business service process of the advertising marketing link and the advertising delivery link.

[0135] For example, in the advertising and marketing phase, feedback information including industry analysis, competitive product analysis, marketing ideas, etc. can be provided in the pre-sales phase. Feedback information including sentiment analysis, intent analysis, and recommended language solutions can be provided in the mid-sales phase. Feedback information including marketing subsections, automatic labels, and next marketing plans can be provided in the post-sales phase.

[0136] For example, in the advertising delivery process, marketing suggestions and feedback information can be provided at various stages before, during, and after the delivery, including advertising goal and strategy customization, keyword research and selection, audience delivery settings, bidding and budgeting, advertising creative production, testing and optimization, data reporting and analysis, etc. The advertising marketing intelligent system can realize the comprehensive service capabilities of artificial intelligence multimodality + global knowledge + industry experience + advertising business capabilities (including analysis, insight, delivery, diagnosis, optimization and other functions) + professional small models + large language models based on intelligent agents, and realize the autonomous solution of user needs in the advertising marketing business domain.

[0137] Figure 7 A framework diagram suitable for executing the information interaction method based on a large language model according to an embodiment of the present disclosure is schematically shown.

[0138] like Figure 7As shown, the intelligent system framework 700 can be applied to implement the information interaction method based on the large language model provided according to the embodiment of the present disclosure. The intelligent system framework 700 may include: a software development kit layer, a service layer, a model layer and a storage layer. The software development kit (SDK) layer may include a business encapsulation software development kit module and an artificial intelligence software development kit module. The software development kit layer may be a core carrier for atomic artificial intelligence capabilities and general scenario business capabilities, which facilitates the target object to input demand description information in time and obtain feedback information. The model layer may include various types of deep learning model support such as large language models and open source models, and may also implement model training customization, large model tuning and other functions through the model tuning module. The storage layer is responsible for data persistence of the entire system, including: a vector subsystem, a cache subsystem, a memory subsystem, a meta information subsystem, a business corpus, etc. The intelligent system framework 700 can provide strong support for the intelligent agent to implement the information interaction method provided by the embodiment of the present disclosure.

[0139] Figure 8 The schematic diagram shows an architecture diagram of an advertising and marketing system suitable for executing the embodiments of the present disclosure.

[0140] like Figure 8 As shown, the advertising and marketing system 800 can be applied to execute the information interaction method based on the large language model provided by the embodiment of the present disclosure. The advertising and marketing system 800 may include an interaction layer at the top level, an intelligent agent at the middle level, and a plug-in layer. The interaction layer can realize information interaction with the target object based on the graphical rendering component library and the UI (interface design) interaction protocol. The intelligent agent can realize other functions such as generative content function by interacting with the large language model and the business model. The intelligent agent can also interact with modules such as the authority management control module, the operation and maintenance monitoring module, the intelligent agent strategy evolution module, the small flow control module, and the data ETL (Extract-Transform-Load, also known as the data warehouse) system module to realize plug-in resource calling, data storage and other functions. The intelligent agent can also interact with the marketing domain library, the marketing domain knowledge base, the analysis platform, the business report center, the multimodal intelligent module, and the public domain capability knowledge base based on the plug-ins such as the retrieval enhancement plug-in, the query statement plug-in, the business plug-in cluster, the public domain plug-in, and other plug-ins.

[0141] According to the embodiments of the present disclosure, the advertising and marketing system can integrate various plug-ins based on the middle-stage design. The richer the plug-in capabilities, the higher the intelligence of the intelligent body. The module deployment of the intelligent body can adopt a centralized mode. The intelligent body switches and generates various task execution strategies by switching various types of large language models. And based on the generative task execution strategy, call various plug-in resources to realize the dynamic arrangement and static arrangement of plug-in resources. Flexible switching. At the same time, multiple plug-ins that support task execution strategies generate a composite plug-in for flexible calling. In the process of intelligent interaction with users, the intermediate thinking process, such as the intermediate task execution results, can be displayed to realize the decision-making process in which users participate. At the same time, the data flywheel function can also be realized. The data flywheel function can, for example, include the brain (large language model) and the cerebellum (MoE, Mixture of Experts, mixed expert model), and through business effect feedback information, draw on the idea of ​​"knowledge distillation" to make the decision-making layer intelligent and iterative.

[0142] Fig. 9 The schematic diagram shows an architecture diagram suitable for executing an intelligent agent according to an embodiment of the present disclosure.

[0143] like Fig. 9 As shown, the intelligent body 900 may include an artificial intelligence business service module, a basic capability + business capability module, a planning & reflection module, a memory capability module, a memory intermediate module, an action module, an intelligent body module, an artificial intelligence model capability module, and an underlying business capability module. The memory capability module may include a long-term memory unit and a working memory unit. The planning & reflection module may include a feedback unit, an annoying optimization unit, a task execution strategy unit, and a task decomposition unit. The basic capability + business capability module may include a variety of plug-in resources such as search plug-ins, interface plug-ins, and permission authentication plug-ins. The underlying business capability module may include: functional units such as an optimization center, an effect platform, and a tripartite platform interface for realizing functions such as information display and data call of feedback information. The artificial intelligence model capability module may include functional units such as text processing units, image processing units, video processing units, and multimodal processing models that are constructed based on artificial intelligence algorithms. The artificial intelligence model capability module and the underlying business capability module can realize that each functional unit is automatically scheduled and assembled by the intelligent body module or the action module.

[0144] Fig.10 A schematic diagram is shown which is suitable for executing the evolution of an intelligent agent according to an embodiment of the present disclosure.

[0145] like Fig.10As shown, the intelligent agent can be jointly developed based on R&D personnel, product personnel, operation personnel, and testers, such as the diversified plug-in resources of R&D personnel, such as artificial intelligence plug-ins, business service plug-in capabilities, etc. Product personnel can provide marketing-related knowledge. Operation personnel can provide industry methods. Input industry experience to the intelligent agent. Testers can provide business test examples. The intelligent agent can provide advertising marketing business service information to users, and can iteratively optimize technical solutions based on user interaction information, and provide it to the data flywheel module, the group intelligence module, and the domain experience module to iteratively optimize the functions of the intelligent agent.

[0146] Fig.11 The following schematically shows a principle diagram suitable for executing agent evolution according to another embodiment of the present disclosure.

[0147] like Fig.11 As shown, the agent can provide the business service module with a task execution strategy, and can obtain the strategy execution result based on the task execution strategy. For example, the business service module can transmit the personalized solution uploaded by the target object to the personalized solution library and the marketing solution library based on the platform arrangement solution generated by the agent to enrich the knowledge information used to optimize the model, and can also obtain public professional knowledge through the public domain knowledge library. The post-training module can obtain knowledge information from any one or more of the personalized solution library, the marketing solution library and the public domain knowledge library through the knowledge collection module. The post-training module can strengthen marketing knowledge through the interactive information provided by the agent, and can also optimize the large language model based on the collected information, and can perform knowledge distillation on the hybrid expert model 1, the hybrid expert model 2, etc. based on the knowledge distillation method, thereby improving the data processing capability of the hybrid expert model.

[0148] This allows the intelligent agent to evolve from the traditional program development mode to the teaching intelligent agent mode, making the intelligent agent more and more powerful. Through the reflection function of the intelligent agent, combined with the plug-in resources, knowledge resources, examples and other information provided by various relevant personnel, the intelligent agent can be continuously "educated", the intelligent agent's logical thinking paradigm can be continuously optimized, the knowledge scope can be expanded, and the intelligent agent's intelligence can be continuously evolved to expand new business value.

[0149] Fig.12 The flowchart of the information interaction method based on the large language model according to another embodiment of the present disclosure is schematically shown.

[0150] like Fig.12 As shown, the information interaction method based on the large language model includes operations S1210 to S1230.

[0151] In operation S1210, in response to the requirement description information input by the target object, at least one of the plurality of agents is used to process the requirement description information to obtain feedback information.

[0152] In operation S1220, at least two of the multiple agents are utilized to implement at least one information interaction based on the feedback information to generate interaction feedback information.

[0153] In operation S1230, interactive feedback information is displayed.

[0154] According to the embodiments of the present disclosure, the intelligent agent may include a device suitable for implementing information processing functions based on deep learning algorithms, for example, it may include software intelligent agents such as virtual devices constructed based on deep learning algorithms, or it may also include hardware intelligent agents such as chips with information processing functions. The embodiments of the present disclosure do not limit the specific structure or type of the intelligent agent, as long as it can perform information processing functions.

[0155] According to an embodiment of the present disclosure, an intelligent agent is adapted to perform the following operations: in response to acquiring demand description information, processing the demand description information using a large language model to obtain a task execution strategy that matches the demand description information, wherein the task execution strategy includes a plurality of demand task attributes, and at least one of the plurality of demand task attributes matches the demand intent represented by the demand description information; determining a strategy execution result related to the task execution strategy according to the demand task attributes in the task execution strategy; and generating feedback information according to the strategy execution result, and displaying the feedback information;

[0156] According to the embodiments of the present disclosure, by having at least two of the multiple intelligent agents interact with each other to generate interactive feedback information, the multiple intelligent agents can interact with each other to realize the target object's demand intention and promote the execution of the required tasks. This can promote task execution by simulating human interaction and thinking abilities without human intervention, so that the generated interactive feedback information can facilitate the display of the task execution process and reasoning process, improve the readability and comprehensibility of the acquired data, and thus enhance the user experience.

[0157] According to an embodiment of the present disclosure, determining a policy execution result related to a task execution strategy based on the requirement task attributes in the task execution strategy includes: determining a completed requirement task attribute that has been processed from multiple requirement task attributes of the current task execution strategy; using a large language model to process the completed requirement task attribute and the task execution results associated with the completed requirement task attribute to obtain the requirement task attribute to be processed; updating the current task execution strategy based on the requirement task attribute to be processed to obtain a new task execution strategy; and determining the policy execution result based on the new task execution strategy.

[0158] According to an embodiment of the present disclosure, at least one of the multiple agents can be used to implement the information interaction method based on the large language model provided by the embodiment of the present disclosure, and the applicant will not elaborate on it here.

[0159] Fig.13 An application scenario diagram of an information interaction method based on a large language model according to another embodiment of the present disclosure is schematically shown.

[0160] like Fig.13 The application scenario shown may include an interactive interface 1300. The target object 1310 may input the requirement description information 1311 "How is the solution performing?", and the intelligent agent 1320 may obtain the first feedback information 1321 "The data analysis of the solution in the past three months is xxx" based on the requirement description information 1311. The intelligent agent 1330 may execute the information interaction method provided by the embodiment of the present disclosure based on the first feedback information 1321 as the requirement description information, and obtain the first interactive feedback information 1331 "The data analysis is represented by a data table". The intelligent agent 1320 may execute the information interaction method provided by the embodiment of the present disclosure based on the first interactive feedback information 1331 as the requirement description information, and obtain the second interactive feedback information 1322.

[0161] Fig.14 A block diagram of an information interaction device based on a large language model according to an embodiment of the present disclosure is schematically shown.

[0162] like Fig.14 As shown, the information interaction device 1400 based on the large language model may include: a task execution strategy acquisition module 1410 , a strategy execution result determination module 1420 and a feedback module 1430 .

[0163] The task execution strategy acquisition module 1410 is used to respond to the acquisition of requirement description information, process the requirement description information using a large language model, and obtain a task execution strategy that matches the requirement description information, wherein the task execution strategy includes multiple requirement task attributes, and at least one of the multiple requirement task attributes matches the requirement intent represented by the requirement description information.

[0164] The policy execution result determination module 1420 is used to determine the policy execution result related to the task execution policy according to the required task attributes in the task execution policy.

[0165] The feedback module 1430 is used to generate feedback information according to the policy execution result and display the feedback information.

[0166] According to an embodiment of the present disclosure, the policy execution result determination module includes: a first determination submodule, a pending demand task attribute acquisition submodule, an update submodule and a policy execution result determination submodule.

[0167] The first determination submodule is used to determine the completed required task attributes from the multiple required task attributes of the current task execution strategy.

[0168] The submodule for obtaining attributes of pending demand tasks is used to use the large language model to process the attributes of completed demand tasks and the task execution results associated with the completed demand task attributes to obtain the attributes of pending demand tasks.

[0169] The update submodule is used to update the current task execution strategy according to the attributes of the task to be processed to obtain a new task execution strategy.

[0170] The strategy execution result determination submodule is used to determine the strategy execution result according to the new task execution strategy.

[0171] According to an embodiment of the present disclosure, the updating submodule includes a first updating unit.

[0172] The first updating unit is used to update the completed demand task attributes in the current task execution strategy according to the pending demand task attributes.

[0173] According to an embodiment of the present disclosure, the updating submodule includes a second updating unit.

[0174] The second updating unit is used to update the unfinished demand task attributes that have not been processed in the current task execution strategy according to the to-be-processed demand task attributes.

[0175] According to an embodiment of the present disclosure, a new task execution strategy includes new execution dependencies between required task attributes.

[0176] According to an embodiment of the present disclosure, the strategy execution result determination submodule includes: a first target requirement task determination unit and a first task execution result acquisition unit.

[0177] The first target requirement task determination unit is used to determine the first target requirement task according to the first target requirement task attribute and the first dependent task execution result associated with the first dependent requirement task attribute, wherein the new task execution strategy includes the first target requirement task attribute and the first dependent requirement task attribute, and there is an execution dependency relationship between the first target requirement task attribute and the first dependent requirement task attribute.

[0178] The first task execution result obtaining unit is used to call the first target service resource associated with the first target requirement task attribute to execute the first target requirement task, and obtain the task execution result associated with the first target requirement task attribute. The policy execution result includes the task execution result associated with the first target requirement task attribute.

[0179] According to an embodiment of the present disclosure, the first target requirement task attribute is characterized based on a prompt template.

[0180] According to an embodiment of the present disclosure, the first target requirement task determining unit includes: a first updating subunit and a first target requirement task obtaining unit.

[0181] The first updating subunit is used to update the first target requirement task attribute according to the first dependent task execution result associated with the first dependent requirement task attribute to obtain the first target task prompt information.

[0182] The first target requirement task obtaining unit is used to use the large language model to process the first target task prompt information to obtain the first target requirement task.

[0183] According to an embodiment of the present disclosure, the task execution strategy also includes execution dependency relationships between multiple required task attributes.

[0184] According to an embodiment of the present disclosure, the strategy execution result determination module includes: a second target requirement task acquisition submodule and a second task execution result acquisition submodule.

[0185] The second target requirement task acquisition submodule is used to generate the second target requirement task according to the second target requirement task attributes and the task execution results associated with the second dependent requirement task attributes, wherein the task execution strategy includes the second dependent requirement task attributes and the second target requirement task attributes, and there is an execution dependency relationship between the second dependent requirement task attributes and the second target requirement task attributes.

[0186] The second task execution result obtaining submodule is used to call the second target service resource associated with the second target requirement task attribute to execute the second target requirement task, and obtain the task execution result associated with the second target requirement task attribute. The policy execution result includes the task execution result associated with the second target requirement task attribute.

[0187] According to an embodiment of the present disclosure, the second target requirement task attribute is characterized based on a prompt template.

[0188] According to an embodiment of the present disclosure, the second target requirement task obtaining submodule includes: a second target task prompt information obtaining unit and a second target requirement task obtaining unit.

[0189] The second target task prompt information obtaining unit is used to update the second target requirement task attribute according to the task execution result associated with the second dependency requirement task attribute to obtain the second target task prompt information.

[0190] The second target requirement task obtaining unit is used to use the large language model to process the second target task prompt information to obtain the second target requirement task.

[0191] According to an embodiment of the present disclosure, the first target requirement task attribute or the second target requirement task attribute includes a target service resource identifier, and the target service resource identifier is suitable for characterizing the association relationship between the first target requirement task attribute and the first target service resource, or is suitable for characterizing the association relationship between the second target requirement task attribute and the second target service resource.

[0192] According to an embodiment of the present disclosure, the information interaction device based on the large language model also includes a first display module.

[0193] The first display module is used to display the target task execution result, wherein the target task execution result includes: the task execution result associated with the first target requirement task attribute; or the task execution result associated with the second target requirement task attribute.

[0194] According to an embodiment of the present disclosure, the task execution strategy acquisition module includes: a multi-level demand intention acquisition sub-module and a task execution strategy determination sub-module.

[0195] The multi-level demand intention acquisition submodule is used to use a large language model to process the demand description information and obtain multi-level demand intentions.

[0196] The task execution strategy determination submodule is used to determine the task execution strategy based on multi-level demand intentions.

[0197] According to an embodiment of the present disclosure, the multi-level demand intention obtaining submodule includes a first demand intention obtaining unit.

[0198] The first demand intention obtaining unit is used to process the demand description information and the demand intention of the previous level with a large language model for the current level to obtain the demand intention of the current level, wherein the demand intention of the first level in the multi-level demand intention is obtained by processing the demand description information with a large language model.

[0199] According to an embodiment of the present disclosure, the multi-level demand intention obtaining submodule includes a second demand intention obtaining unit.

[0200] The second demand intention obtaining unit is used to use a large language model to process the demand description information and the associated query information queried based on the demand description information to obtain multi-level demand intentions.

[0201] According to an embodiment of the present disclosure, the information interaction device based on the large language model also includes: a second display module and a demand intention determination module.

[0202] The second display module is used to display at least one demand intention among the multi-level demand intentions.

[0203] The demand intention determination module is used to determine the demand intention associated with the selection operation in response to the selection operation for at least one demand intention.

[0204] According to an embodiment of the present disclosure, the task execution strategy determination submodule includes a task execution strategy determination unit.

[0205] The task execution strategy determination unit is used to determine the task execution strategy according to the demand intention associated with the selection operation in the multi-level demand intention.

[0206] Fig.15 A block diagram of an information interaction device based on a large language model according to another embodiment of the present disclosure is schematically shown.

[0207] like Fig.15 As shown, the information interaction device 1500 based on the large language model includes: a feedback information obtaining module 1510, an interactive feedback information generating module 1520 and an interactive feedback information display module 1530.

[0208] The feedback information acquisition module 1510 is used to respond to the demand description information input by the target object, use at least one intelligent agent among multiple intelligent agents to process the demand description information, and obtain feedback information. The intelligent agent is suitable for performing the following operations: in response to obtaining the demand description information, use a large language model to process the demand description information, and obtain a task execution strategy that matches the demand description information, wherein the task execution strategy includes multiple demand task attributes, and at least one of the multiple demand task attributes matches the demand intent represented by the demand description information; determine the strategy execution result related to the task execution strategy based on the demand task attributes in the task execution strategy; and generate feedback information based on the strategy execution result, and display the feedback information.

[0209] The interactive feedback information generating module 1520 is used to utilize at least two of the multiple agents to implement at least one information interaction based on feedback information to generate interactive feedback information.

[0210] The interactive feedback information display module 1530 is used to display the interactive feedback information.

[0211] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0212] According to an embodiment of the present disclosure, an electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method provided according to the embodiment of the present disclosure.

[0213] According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable a computer to execute a method provided according to an embodiment of the present disclosure.

[0214] According to an embodiment of the present disclosure, a computer program product includes a computer program, and when the computer program is executed by a processor, the method provided according to the embodiment of the present disclosure is implemented.

[0215] Fig.16 A block diagram of an electronic device suitable for implementing an information interaction method based on a large language model according to an embodiment of the present disclosure is schematically shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0216] like Fig.16 As shown, the device 1600 includes a computing unit 1601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1602 or a computer program loaded from a storage unit 1608 into a random access memory (RAM) 1603. In the RAM 1603, various programs and data required for the operation of the device 1600 can also be stored. The computing unit 1601, the ROM 1602, and the RAM 1603 are connected to each other via a bus 1604. An input / output (I / O) interface 1605 is also connected to the bus 1604.

[0217] A number of components in the device 1600 are connected to the I / O interface 1605, including: an input unit 1606, such as a keyboard, a mouse, etc.; an output unit 1607, such as various types of displays, speakers, etc.; a storage unit 1608, such as a disk, an optical disk, etc.; and a communication unit 1609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1609 allows the device 1600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0218] The computing unit 1601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1601 performs the various methods and processes described above, such as an information interaction method based on a large language model. For example, in some embodiments, the information interaction method based on a large language model may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 1608. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1600 via the ROM 1602 and / or the communication unit 1609. When the computer program is loaded into the RAM 1603 and executed by the computing unit 1601, one or more steps of the information interaction method based on the large language model described above may be performed. Alternatively, in other embodiments, the computing unit 1601 may be configured to execute the information interaction method based on the large language model in any other appropriate manner (for example, by means of firmware).

[0219] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0220] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0221] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0222] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0223] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0224] A computer system may include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises through computer programs running on respective computers and having a client-server relationship to each other. The server may be a cloud server, a server in a distributed system, or a server combined with a blockchain.

[0225] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0226] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An information interaction method based on a large language model, comprising: In response to acquiring the requirement description information, the requirement description information is processed using a large language model to obtain a task execution strategy matching the requirement description information, the task execution strategy including a plurality of requirement task attributes and execution dependencies between the plurality of requirement task attributes, at least one of the plurality of requirement task attributes matching the requirement intent represented by the requirement description information; Determining a policy execution result related to the task execution strategy according to the required task attributes in the task execution strategy; and Generate feedback information according to the strategy execution result, and display the feedback information; Determining a strategy execution result related to the task execution strategy according to the required task attributes in the task execution strategy includes: Determining a completed required task attribute from the multiple required task attributes of the current task execution strategy; Using a large language model to process the completion demand task attributes and the task execution results associated with the completion demand task attributes, to obtain the demand task attributes to be processed; Update the current task execution strategy according to the attributes of the pending demand task to obtain a new task execution strategy; and According to the new task execution strategy, the strategy execution result is determined.

2. The method according to claim 1, wherein: The updating of the current task execution strategy according to the attributes of the pending demand task comprises: Update the completed demand task attribute in the current task execution strategy according to the pending demand task attribute; and / or The unfinished demand task attributes that have not been processed in the current task execution strategy are updated according to the to-be-processed demand task attributes.

3. The method according to claim 1, wherein: The new task execution strategy includes execution dependencies between new required task attributes; Wherein, determining the strategy execution result according to the new task execution strategy includes: Determine the first target requirement task according to the first target requirement task attribute and the first dependent task execution result associated with the first dependent requirement task attribute, wherein the new task execution strategy includes the first target requirement task attribute and the first dependent requirement task attribute, and there is an execution dependency relationship between the first target requirement task attribute and the first dependent requirement task attribute; Call the first target service resource associated with the first target requirement task attribute to execute the first target requirement task, and obtain the task execution result associated with the first target requirement task attribute, wherein the policy execution result includes the task execution result associated with the first target requirement task attribute.

4. The method according to claim 3, wherein the first target requirement task attribute is characterized based on a prompt template; in, Determining the first target requirement task according to the first target requirement task attribute and the first dependent task execution result associated with the first dependent requirement task attribute includes: According to the first dependent task execution result associated with the first dependent demand task attribute, the first target demand task attribute is updated to obtain first target task prompt information; and The first target task prompt information is processed using a large language model to obtain the first target requirement task.

5. The method according to claim 1, in, Determining a strategy execution result related to the task execution strategy according to the required task attributes in the task execution strategy includes: Generate a second target requirement task according to the second target requirement task attribute and the task execution result associated with the second dependent requirement task attribute, wherein the task execution strategy includes the second dependent requirement task attribute and the second target requirement task attribute, and there is an execution dependency relationship between the second dependent requirement task attribute and the second target requirement task attribute; and Call the second target service resource associated with the second target requirement task attribute to execute the second target requirement task, and obtain the task execution result associated with the second target requirement task attribute, wherein the policy execution result includes the task execution result associated with the second target requirement task attribute.

6. The method according to claim 5, wherein: The second target requirement task attribute is characterized based on a prompt template; The generating of the second target requirement task according to the second target requirement task attribute and the task execution result associated with the second dependent requirement task attribute includes: According to the task execution result associated with the second dependent requirement task attribute, updating the second target requirement task attribute to obtain second target task prompt information; and The second target task prompt information is processed using a large language model to obtain the second target requirement task.

7. The method according to claim 3, wherein: The first target requirement task attribute includes a target service resource identifier, and the target service resource identifier is suitable for representing an association relationship between the first target requirement task attribute and the first target service resource.

8. The method according to claim 5, wherein: The second target requirement task attribute includes a target service resource identifier, and the target service resource identifier is suitable for representing the association relationship between the second target requirement task attribute and the second target service resource.

9. The method according to claim 3, further comprising: Display the target task execution result, wherein the target task execution result includes: the task execution result associated with the first target requirement task attribute.

10. The method according to claim 5, further comprising: Display the target task execution result, wherein the target task execution result includes the task execution result associated with the second target requirement task attribute.

11. The method according to claim 1, wherein: The using of a large language model to process the requirement description information to obtain a task execution strategy matching the requirement description information comprises: Processing the demand description information using a large language model to obtain multi-level demand intentions; and Determine the task execution strategy according to the multi-level demand intentions.

12. The method according to claim 11, wherein: The use of a large language model to process the demand description information to obtain a multi-level demand intent includes: For the current level, the large language model is used to process the demand description information and the demand intention of the previous level to obtain the demand intention of the current level, wherein the demand intention of the first level among the multi-level demand intentions is obtained by processing the demand description information using the large language model.

13. The method according to claim 11, wherein: The use of a large language model to process the demand description information to obtain a multi-level demand intent includes: The demand description information is processed using a large language model, and the associated query information queried based on the demand description information is used to obtain multi-level demand intentions.

14. The method according to any one of claims 11 to 13, before determining the task execution strategy according to the multi-level demand intentions, the method comprises: Displaying at least one of the multi-level demand intentions; as well as In response to a selection operation for at least one of the demand intentions, determining the demand intention associated with the selection operation; Wherein, determining the task execution strategy according to the multi-level demand intentions includes: The task execution strategy is determined according to the demand intention associated with the selection operation in the multi-level demand intentions.

15. An information interaction method based on a large language model, comprising: In response to the requirement description information input by the target object, at least one of the multiple agents is used to process the requirement description information to obtain feedback information, and the agent is suitable for performing the following operations: In response to acquiring the requirement description information, the requirement description information is processed using a large language model to obtain a task execution strategy matching the requirement description information, wherein the task execution strategy includes a plurality of requirement task attributes and execution dependencies between the plurality of requirement task attributes, and at least one of the plurality of requirement task attributes matches the requirement intent represented by the requirement description information; Determining a policy execution result related to the task execution strategy according to the required task attributes in the task execution strategy; and Generate feedback information according to the strategy execution result, and display the feedback information; Using at least two of the plurality of intelligent agents to implement at least one information interaction based on the feedback information to generate interactive feedback information; and displaying the interactive feedback information; Wherein, determining the policy execution result related to the task execution policy according to the required task attributes in the task execution policy includes: Determining a completed required task attribute from the multiple required task attributes of the current task execution strategy; Using a large language model to process the completion demand task attributes and the task execution results associated with the completion demand task attributes, to obtain the demand task attributes to be processed; Update the current task execution strategy according to the attributes of the pending demand task to obtain a new task execution strategy; and According to the new task execution strategy, the strategy execution result is determined.

16. An information interaction device based on a large language model, comprising: A task execution strategy acquisition module is used to, in response to acquiring the requirement description information, process the requirement description information using a large language model to obtain a task execution strategy that matches the requirement description information, wherein the task execution strategy includes a plurality of requirement task attributes and execution dependencies between the plurality of requirement task attributes, and at least one of the plurality of requirement task attributes matches the requirement intent represented by the requirement description information; a strategy execution result determination module, configured to determine a strategy execution result related to the task execution strategy according to the required task attributes in the task execution strategy; and A feedback module, used to generate feedback information according to the strategy execution result and display the feedback information; Wherein, the strategy execution result determination module includes: A first determination submodule is used to determine the completed required task attribute from the multiple required task attributes of the current task execution strategy; The submodule for obtaining attributes of the required task to be processed is used to process the attributes of the completed required task and the task execution results associated with the attributes of the completed required task using the large language model to obtain the attributes of the required task to be processed; An updating submodule, used to update the current task execution strategy according to the attributes of the pending demand task to obtain a new task execution strategy; and The strategy execution result determination submodule is used to determine the strategy execution result according to the new task execution strategy.

17. The device according to claim 16, wherein: The updating submodule comprises: A first updating unit is used to update the completion requirement task attribute in the current task execution strategy according to the pending requirement task attribute; and / or The second updating unit is used to update the unfinished demand task attributes that have not been processed in the current task execution strategy according to the to-be-processed demand task attributes.

18. The device according to claim 16, wherein: The new task execution strategy includes execution dependencies between new required task attributes; The strategy execution result determination submodule includes: A first target requirement task determination unit is used to determine the first target requirement task according to the first target requirement task attribute and the first dependent task execution result associated with the first dependent requirement task attribute, wherein the new task execution strategy includes the first target requirement task attribute and the first dependent requirement task attribute, and there is an execution dependency relationship between the first target requirement task attribute and the first dependent requirement task attribute; The first task execution result obtaining unit is used to call the first target service resource associated with the first target requirement task attribute to execute the first target requirement task, and obtain the task execution result associated with the first target requirement task attribute. The policy execution result includes the task execution result associated with the first target requirement task attribute.

19. The apparatus of claim 18, wherein the first target requirement task attribute is characterized based on a prompt template; in, The first target requirement task determination unit includes: A first updating subunit is configured to update the first target requirement task attribute according to a first dependent task execution result associated with the first dependent requirement task attribute, and obtain first target task prompt information; and The first target requirement task obtaining unit is used to use the large language model to process the first target task prompt information to obtain the first target requirement task.

20. The device according to claim 16, in, The strategy execution result determination module includes: a second target requirement task acquisition submodule, for generating a second target requirement task according to a second target requirement task attribute and a task execution result associated with a second dependent requirement task attribute, wherein the task execution strategy includes the second dependent requirement task attribute and the second target requirement task attribute, and there is an execution dependency relationship between the second dependent requirement task attribute and the second target requirement task attribute; and The second task execution result obtaining submodule is used to call the second target service resource associated with the second target requirement task attribute to execute the second target requirement task, and obtain the task execution result associated with the second target requirement task attribute. The policy execution result includes the task execution result associated with the second target requirement task attribute.

21. The device according to claim 20, wherein: The second target requirement task attribute is characterized based on a prompt template; Wherein, the second target requirement task acquisition submodule includes: A second target task prompt information obtaining unit is configured to update the second target requirement task attribute according to the task execution result associated with the second dependency requirement task attribute to obtain the second target task prompt information; and The second target requirement task obtaining unit is used to use the large language model to process the second target task prompt information to obtain the second target requirement task.

22. The device according to claim 18, wherein The first target requirement task attribute includes a target service resource identifier, and the target service resource identifier is suitable for representing an association relationship between the first target requirement task attribute and the first target service resource.

23. The device according to claim 20, wherein: The second target requirement task attribute includes a target service resource identifier, and the target service resource identifier is suitable for representing the association relationship between the second target requirement task attribute and the second target service resource.

24. The apparatus of claim 18, further comprising: The first display module is used to display the target task execution result, wherein the target task execution result includes the task execution result associated with the first target requirement task attribute.

25. The apparatus of claim 20, further comprising: The first display module is used to display the target task execution result, wherein the target task execution result includes the task execution result associated with the second target requirement task attribute.

26. The device according to claim 16, wherein: The task execution strategy acquisition module includes: A multi-level demand intention obtaining submodule is used to process the demand description information using a large language model to obtain a multi-level demand intention; and The task execution strategy determination submodule is used to determine the task execution strategy according to the multi-level demand intentions.

27. The device according to claim 26, wherein: The multi-level demand intention acquisition submodule includes: The first demand intention obtaining unit is used to process the demand description information and the demand intention of the previous level with the large language model for the current level to obtain the demand intention of the current level, wherein the demand intention of the first level among the multi-level demand intentions is obtained by processing the demand description information with the large language model.

28. The device according to claim 26, wherein The multi-level demand intention acquisition submodule includes: The second demand intention obtaining unit is used to process the demand description information using a large language model, and obtain multi-level demand intentions based on the associated query information queried based on the demand description information.

29. The apparatus according to any one of claims 26 to 28, further comprising: A second display module, used to display at least one of the multi-level demand intentions; as well as A demand intention determination module, configured to determine the demand intention associated with the selection operation in response to a selection operation for at least one of the demand intentions; The task execution strategy determination submodule includes: The task execution strategy determination unit is used to determine the task execution strategy according to the demand intention associated with the selection operation in the multi-level demand intention.

30. An information interaction device based on a large language model, comprising: The feedback information obtaining module is used to respond to the requirement description information input by the target object, use at least one of the multiple agents to process the requirement description information, and obtain feedback information, wherein the agent is suitable for performing the following operations: In response to acquiring the requirement description information, the requirement description information is processed using a large language model to obtain a task execution strategy matching the requirement description information, wherein the task execution strategy includes a plurality of requirement task attributes and execution dependencies between the plurality of requirement task attributes, and at least one of the plurality of requirement task attributes matches the requirement intent represented by the requirement description information; Determining a policy execution result related to the task execution strategy according to the required task attributes in the task execution strategy; and Generate feedback information according to the strategy execution result, and display the feedback information; an interactive feedback information generating module, configured to utilize at least two of the plurality of intelligent agents to implement at least one information interaction based on the feedback information, and generate interactive feedback information; and An interactive feedback information display module, used to display the interactive feedback information; The agent is further configured to perform the following operations: Determining a completed required task attribute from the multiple required task attributes of the current task execution strategy; Using a large language model to process the completion demand task attributes and the task execution results associated with the completion demand task attributes, to obtain the demand task attributes to be processed; Update the current task execution strategy according to the attributes of the pending demand task to obtain a new task execution strategy; and According to the new task execution strategy, the strategy execution result is determined.

31. 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 executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 15.

32. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 15.

33. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 15.

Images