Task processing method, computing device, and computer-readable storage medium

Abstract

The present application discloses a task processing method, a computing device, and a readable storage medium, including: decomposing a target task into multiple subtasks based on a deep learning model; evaluating the multiple subtasks to determine a first evaluation value of the multiple subtasks; if the first evaluation value is lower than a first preset value, then decomposing the target task again based on the deep learning model until the first evaluation value of the multiple subtasks obtained by decomposing the target task is not lower than the first preset value; if the first evaluation value is not lower than the first preset value, then executing corresponding processing strategies for the subtasks according to the processing difficulty of the subtasks. In this way, by evaluating the task decomposition process to optimize the decomposition of the target task and matching the optimal processing strategy according to the processing difficulty of each subtask after decomposition, the resource utilization efficiency and the automation and intelligence levels of task decomposition are improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and particularly to a task processing method, a computing device, and a computer-readable storage medium. Background Art

[0002] In recent years, generative large models (such as GPT, etc.) have made remarkable progress in the field of natural language processing, demonstrating excellent versatility and generative capabilities, and can complete various tasks such as text generation, language translation, code generation, etc. However, when faced with complex tasks, these models expose certain limitations, such as unreasonable task decomposition, low execution efficiency, or resource waste, etc. Summary of the Invention

[0003] The purpose of this application is to provide a task processing method, a computing device, and a computer-readable storage medium, which optimize the decomposition of the target task by evaluating the task decomposition process, and match the optimal processing strategy according to the processing difficulty of each subtask after decomposition, improving the resource utilization efficiency and the automation and intelligence level of task decomposition.

[0004] To achieve the above purpose:

[0005] In a first aspect, an embodiment of this application provides a task processing method, including the following steps:

[0006] Decompose the target task into multiple subtasks based on a deep learning model;

[0007] Evaluate multiple subtasks to determine the first evaluation value of the multiple subtasks;

[0008] If the first evaluation value is lower than the first preset value, then decompose the target task again based on the deep learning model until the first evaluation value of the multiple subtasks obtained by decomposing the target task is not lower than the first preset value;

[0009] If the first evaluation value is not lower than the first preset value, then execute the corresponding processing strategy for the subtasks according to the processing difficulty of the subtasks.

[0010] In an embodiment, evaluating multiple subtasks includes:

[0011] Evaluate multiple subtasks based on a task decomposition evaluation criterion, where the task decomposition evaluation criterion includes at least one of logical coherence, executability, rationality of subtask granularity, coverage integrity, input-output consistency, subtask relevance, and workload balance.

[0012] In an embodiment, executing the corresponding task processing strategy for the subtasks according to the processing difficulty of the subtasks includes:

[0013] Evaluate the processing difficulty of the subtasks to determine the second evaluation value;

[0014] If the second evaluation value is lower than the second preset value, the subtask is processed based on multiple deep learning models.

[0015] In one embodiment, evaluating the processing difficulty of the subtask includes:

[0016] Evaluating the processing difficulty of the subtask based on a task difficulty evaluation criterion, where the task difficulty evaluation criterion includes at least one of task complexity, estimated resource consumption, data dependency, depth of domain knowledge, and result uncertainty.

[0017] In one embodiment, processing the subtask based on multiple deep learning models includes:

[0018] Processing the subtask based on multiple deep learning models to obtain multiple deep learning model processing results;

[0019] Evaluating the multiple deep learning model processing results respectively based on a preset evaluation criterion;

[0020] Fusing the multiple deep learning model processing results based on the evaluation results to obtain the processing result of the subtask.

[0021] In one embodiment, before processing the subtask based on multiple deep learning models, the task processing method further includes:

[0022] Evaluating the input data of the subtask based on a preset evaluation criterion;

[0023] If the input data of the subtask does not meet the input judgment criterion, adjusting the input data of the subtask until it meets the input judgment criterion.

[0024] In one embodiment, the preset evaluation criterion includes at least one of description matching degree, structural integrity, input accuracy, content consistency, and semantic coherence.

[0025] In one embodiment, the evaluation result is the evaluation score of the multiple deep learning model processing results. Fusing the multiple deep learning model processing results based on the evaluation results to obtain the processing result of the subtask includes:

[0026] Respectively extracting the highest score corresponding to each evaluation criterion in the preset evaluation criterion from the evaluation scores of the multiple deep learning model processing results as the model processing result corresponding to the evaluation criterion;

[0027] Fusing the model processing results corresponding to each evaluation criterion to obtain the processing result of the subtask.

[0028] In a second aspect, an embodiment of the present application provides a computing device, specifically including:

[0029] A processor;

[0030] A memory for storing processor-executable instructions;

[0031] Wherein, the processor is configured to execute instructions for performing the task processing method as described in the first aspect.

[0032] In a third aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored therein. When the instructions in the computer-readable storage medium are executed by a processor of a computing device, the computing device can implement the task processing method as described in the first aspect.

[0033] The task processing method, computing device, and readable storage medium provided by the embodiments of the present application include: decomposing a target task into multiple subtasks based on a deep learning model; evaluating the multiple subtasks to determine a first evaluation value of the multiple subtasks; if the first evaluation value is lower than a first preset value, then decomposing the target task again based on the deep learning model until the first evaluation value of the multiple subtasks obtained by decomposing the target task is not lower than the first preset value; if the first evaluation value is not lower than the first preset value, then execute corresponding processing strategies for the subtasks according to the processing difficulty of the subtasks. In this way, by evaluating the task decomposition process to optimize the decomposition of the target task and matching the optimal processing strategy according to the processing difficulty of each subtask after decomposition, the resource utilization efficiency and the automation and intelligence levels of task decomposition are improved. Description of the Drawings

[0034] Figure 1 It is a schematic flowchart of the task processing method provided by an embodiment of the present invention.

[0035] Figure 2 It is a specific schematic flowchart in the task processing method provided by an embodiment of the present invention.

[0036] Figure 3 It is a schematic structural diagram of the computing device provided by an embodiment of the present invention. Detailed Embodiments

[0037] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0038] It should be noted that in this text, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of another identical element in the process, method, article or device comprising such element. In addition, components, features, and elements with the same name in different embodiments of this application may have the same meaning or may have different meanings, and their specific meanings need to be determined based on their explanations in the specific embodiments or further in combination with the context of the specific embodiments.

[0039] It should be understood that although the terms first, second, third, etc. may be used herein to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of this text, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein can be interpreted as "when" or "upon" or "in response to determining". Furthermore, as used herein, the singular forms "a", "an" and "the" are also intended to include the plural forms unless the context indicates otherwise. It should be further understood that the terms "comprising", "including" indicate the presence of the stated features, steps, operations, elements, components, items, kinds, and / or groups, but do not exclude the presence, occurrence or addition of one or more other features, steps, operations, elements, components, items, kinds, and / or groups. The terms "or" and "and / or" as used herein are interpreted as inclusive, or meaning any one or any combination. Thus, "A, B or C" or "A, B and / or C" means "any of the following: A; B; C; A and B; A and C; B and C; A, B and C". An exception to this definition only occurs when the combination of elements, functions, steps or operations is inherently mutually exclusive in some way.

[0040] It should be understood that although the steps in the flowchart in the embodiments of the present application are sequentially shown according to the indication of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit and can be executed in other orders. Moreover, at least a part of the steps in the figure may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0041] It should be noted that in this article, step codes such as S1 and S2 are used. The purpose is to more clearly and briefly express the corresponding content and do not constitute a substantial limitation in order. Those skilled in the art may execute S2 first and then S1 during specific implementation, etc., but these should all be within the protection scope of the present application.

[0042] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0043] In the subsequent description, suffixes such as "module", "component" or "unit" used to represent elements are only for the convenience of the description of the present application and have no specific meaning in themselves. Therefore, "module", "component" or "unit" can be used interchangeably.

[0044] Most of the existing task decomposition methods are static and fail to make dynamic adjustments according to the real-time status of subtasks. At the same time, the quality of input data and the output integration strategy are crucial for the task execution effect. However, traditional methods often neglect the evaluation and optimization of the integrity and accuracy of the input, and lack the ability to flexibly adjust when facing input data that does not meet the expected requirements, which affects the accuracy and consistency of the output results.

[0045] In addition, for the fusion of different answers in complex tasks, most existing methods use a single algorithm and cannot make full use of the advantages of each answer, which limits the potential of generative models in dealing with complex tasks. Therefore, in order to better solve complex tasks, it is necessary to develop a more intelligent task decomposition mechanism, optimize the input data processing process, and improve the output fusion strategy to enhance the ability of generative large models to handle complex tasks.

[0046] Refer to Figure 1 , to address the above challenges, the embodiments of the present application provide a task processing method. In this embodiment, taking the task processing method applied to a computing device as an example, the task processing method provided in this embodiment includes:

[0047] Step S1: Decompose the target task into multiple subtasks based on a deep learning model.

[0048] Among them, the target task consists of multiple subtasks with dependencies and logic. By decomposing the target into multiple subtasks, the processing difficulty of the target task can be reduced and the execution efficiency can be improved. When decomposing the target task, the generative large model based on deep learning is called to decompose the target task into multiple subtasks with logical association relationships. An independently developed deep learning model can be used, or an existing model can be called, such as the GPT series (Generative Pre-trained Transformer), BERT (Bidirectional Encoder Representations from Transformers), the language model Qwen, etc.

[0049] Step S2: Evaluate multiple subtasks to determine the first evaluation values of the multiple subtasks.

[0050] Among them, multiple subtasks are evaluated through the task decomposition evaluation criteria, and the evaluation content includes but is not limited to the content of the subtasks and the relationships between the subtasks. Through evaluation, the decomposition result of the target task is optimized, so that the complex target task can be scientifically and reasonably disassembled into multiple subtasks, improving the execution efficiency and accuracy of the target task.

[0051] In one implementation, evaluating multiple subtasks includes:

[0052] Evaluating multiple subtasks based on the task decomposition evaluation criteria, where the task decomposition evaluation criteria include at least one of logical coherence, executability, rationality of subtask granularity, coverage integrity, input-output consistency, subtask relevance, and workload balance.

[0053] Optionally, the task decomposition evaluation criteria are used to evaluate the rationality and effectiveness of the generative large model in disassembling complex tasks into subtasks. By setting scientific evaluation indicators, it is ensured that the decomposed subtasks meet high standards in terms of logical coherence, executability, and integrity, thus laying a solid foundation for the smooth execution of subsequent tasks. The task decomposition evaluation criteria include multiple evaluation indicators, and each indicator only targets a specific aspect of the task decomposition performance, that is, it is used to evaluate a certain characteristic of the subtasks or between the subtasks, to ensure the interpretability and operability of the evaluation indicators. By decomposing the complex evaluation task into clear and easy-to-execute independent sub-evaluation tasks through specific evaluation indicators in multiple different dimensions, the accuracy of the evaluation can be improved. On this basis, the scoring ranges of each evaluation indicator are uniformly standardized to facilitate the comparison and comprehensive processing of different evaluation indicators.

[0054] The following describes each evaluation indicator in the task decomposition evaluation criteria.

[0055] Optionally, a logical coherence metric is used to evaluate whether the logical relationships between subtasks are reasonable, ensuring that they align with the overall task objective. This metric checks for clear dependencies between subtasks and whether their order follows a reasonable logical flow, avoiding logical conflicts or inconsistencies.

[0056] The evaluation of the logical coherence metric includes checking whether the dependencies between subtasks are clear, such as one subtask being a prerequisite for another to start, and whether the processing order of each subtask follows a logical flow, e.g., data collection first and then data analysis.

[0057] The scoring range for logical coherence is 0 to 1, where 1 indicates that the logic between the subtasks after decomposing the target task is completely coherent, with no logical loopholes or conflicts found; 0 indicates obvious logical problems that seriously affect the coherence and executability of the task.

[0058] Optionally, an executability metric is used to evaluate whether each subtask has clear inputs and outputs, ensuring that they can be actually executed. This metric checks whether the description of the subtask is clear and specific enough, and whether there are clear input requirements and expected outputs to ensure the practical operability of the task.

[0059] The evaluation of the executability metric includes: checking the clarity and specificity of the subtask description to confirm whether it provides enough details to guide execution; and, confirming whether the input requirements and expected outputs of each subtask are clear, including required resources, tools, data formats, etc.

[0060] The scoring range for the executability metric is 0 to 1, where 1 indicates fully executable, meaning that all subtasks have clear operation guides and expected results; 0 indicates obvious execution obstacles that make the task difficult to implement.

[0061] Optionally, a subtask granularity rationality metric is used to evaluate whether the decomposition of subtasks is appropriate, avoiding being too fine or too coarse. This metric ensures that the task decomposition does not increase unnecessary complexity and management burden due to over - refinement, nor does it affect the efficiency and quality of task execution due to over - coarsening.

[0062] The evaluation of subtask granularity rationality includes comparing the number of subtasks with the overall complexity of the task to judge whether the decomposition is reasonable; and, judging the complexity of individual subtasks to ensure that they are neither too simple nor too complex to be completed independently.

[0063] The scoring range for the rationality of sub-task granularity is 0 to 1, where 1 indicates that the sub-task granularity is completely appropriate and can effectively support task execution; 0 indicates that the granularity is too fine or too coarse, which has a negative impact on task execution.

[0064] Optionally, the coverage integrity metric is used to evaluate whether the decomposed sub-tasks comprehensively cover all the requirements and objectives of the original task, avoid missing task steps, and ensure the integrity of task completion.

[0065] The evaluation content of the coverage integrity metric includes listing all the requirements of the original task and comparing them with the decomposed sub-tasks to check whether there are any key tasks or steps that have not been decomposed. The scoring range for the coverage integrity metric is 0 to 1, where 1 indicates complete coverage.

[0066] Optionally, the input-output consistency metric is used to evaluate whether the inputs and outputs between sub-tasks match and whether the information transfer is smooth, so as to ensure smooth connection between sub-tasks, avoid information gaps or duplicate processing, and improve task processing efficiency.

[0067] The evaluation content of the input-output consistency metric includes checking whether the output of the previous sub-task meets the input requirements of the next sub-task, and confirming the accuracy of data and information transfer between sub-tasks. The scoring range for the input-output consistency metric is 0 to 1, where 1 indicates complete consistency.

[0068] Optionally, the sub-task relevance metric is used to evaluate whether a sub-task directly contributes to the completion of the overall task objective, avoid irrelevant or secondary tasks, and ensure that each sub-task makes a practical contribution to the completion of the overall task.

[0069] The evaluation content of the sub-task relevance metric includes judging the degree of relevance between the objective of the sub-task and the overall task, and identifying whether there are any sub-tasks that are irrelevant to the task objective. The scoring range for the sub-task relevance metric is 0 to 1, where 1 indicates a high degree of relevance.

[0070] Optionally, the workload balance metric is used to evaluate whether the complexity between sub-tasks is balanced, avoid some sub-tasks being too complex or too simple, ensure reasonable task allocation, avoid uneven resource allocation, and ensure the high efficiency of task execution and the rationality of resource utilization.

[0071] The evaluation content of the workload balance metric includes comparing the expected workload and difficulty of each sub-task, and analyzing and confirming whether there is an obvious complexity imbalance between each sub-task. The scoring range for the workload balance metric is 0 to 1, where 1 indicates good complexity balance.

[0072] Exemplarily, assume that the target task is to develop a new mobile application. This task can be decomposed into the following sub-tasks:

[0073] Requirement Analysis: Communicate with the customer to determine the functional requirements and user experience requirements of the application.

[0074] Deployment and Go Live: Release the final version of the application to the app store for users to download.

[0075] Design Prototype: Based on the results of the requirement analysis, create a preliminary design and user interface prototype of the application.

[0076] Technology Selection: Select appropriate technology stacks and tools to implement the functions of the application.

[0077] Coding Implementation: Write code according to the design and technology selection to implement the functions of the application.

[0078] Testing and Optimization: Conduct comprehensive testing on the application and make necessary optimization adjustments based on the feedback.

[0079] For the logical coherence metric, the above 4 subtasks have clear logical dependencies (such as design prototype and technology selection), but 2 subtasks have logical conflicts (for example, there should be no direct association between deployment and go live and requirement analysis). Therefore, the logical coherence score = 4 / 6 = 0.667.

[0080] For the executability metric, evaluate whether each subtask has clear inputs and outputs. It is found that the inputs and outputs of 5 subtasks (such as requirement analysis, design prototype, technology selection, coding implementation, testing and optimization) are clear and executable, while some details of deployment and go live are too abstract and lack clear operation guidelines. Therefore, the executability score = 5 / 6 = 0.833.

[0081] For the rationality of subtask granularity metric, among the 6 subtasks, it is found that the two tasks of "requirement analysis" and "design prototype" are relatively simple and can be combined or further refined; while the workload and complexity of "coding implementation" are much higher than other tasks and seem too complex; the remaining 3 tasks (technology selection, testing and optimization, deployment and go live) have appropriate granularity. Therefore, the subtask granularity rationality score = 3 / 6 = 0.5.

[0082] For the coverage integrity metric, the original tasks had 10 key objectives, including user experience design, performance optimization, security measures, etc. After checking, it is found that the decomposed subtasks cover 9 objectives and one key step regarding user training is missed. Therefore, the coverage integrity score = 9 / 10 = 0.9.

[0083] For the input-output consistency metric, check the input-output matching situation between each pair of subtasks. It is found that the information transfer between 4 pairs of subtasks (such as design prototype and technology selection) is smooth, and there is only an information gap between deployment and go live and the previous requirement analysis link. Therefore, the input-output consistency score = 4 / 5 = 0.8.

[0084] For the subtask relevance metric, analyze whether each subtask directly contributes to achieving the overall goal. It is found that although the subtask of "deployment and go live" is important, in the current decomposition, it is only related to secondary goals, while the other 5 tasks closely revolve around the core goal of application development. Therefore, the subtask relevance score = 5 / 6 = 0.833.

[0085] For the workload balance metric, evaluate the complexity distribution of subtasks. It is found that the subtask of "coding implementation" is overly complex and occupies a large amount of resources; "requirements analysis" and "design prototype" are relatively simple; the other 3 tasks have balanced complexity. Therefore, the workload balance score = 3 / 6 = 0.5.

[0086] Based on the above metric scores (assuming equal weights for each metric), the comprehensive score is calculated as follows:

[0087] The first evaluation value = (4 / 6 + 5 / 6 + 3 / 6 + 9 / 10 + 4 / 5 + 5 / 6 + 3 / 6) / 7 = 0.719.

[0088] Of course, in other embodiments, different weights can also be set according to the metrics that need to be emphasized.

[0089] Step S3: If the first evaluation value is lower than the first preset value, then decompose the target task again based on the deep learning model until the first evaluation value of the multiple subtasks obtained by decomposing the target task is not lower than the first preset value.

[0090] In the above case, assume that the first preset value is 0.8, that is, the first evaluation value is lower than the first preset value. Then decompose the target task again based on the deep learning model and evaluate the multiple subtasks obtained by decomposition again until the first evaluation value of the multiple subtasks is not lower than the first preset value, so as to improve the rationality and scientificity of task decomposition.

[0091] Step S4: If the first evaluation value is not lower than the first preset value, then execute corresponding processing strategies for the subtasks according to the processing difficulty of the subtasks.

[0092] Among them, if the first evaluation value is not lower than the first preset value, it indicates that the decomposition of the target task is completed, and the decomposition process meets the task decomposition evaluation criteria, and the multiple subtasks obtained by decomposition can be processed continuously. The task difficulty dynamic evaluation mechanism enables the system to adjust strategies according to the real-time state of the task, thereby improving the flexibility of complex task processing.

[0093] In one implementation, executing corresponding task processing strategies for the subtasks according to the processing difficulty of the subtasks includes:

[0094] Evaluate the processing difficulty of the subtasks to determine the second evaluation value;

[0095] If the second evaluation value is lower than the second preset value, the subtask is processed based on multiple deep learning models.

[0096] Optionally, first evaluate the processing difficulty of the subtask to determine a quantified second evaluation value. If the second evaluation value is lower than the second preset value, it indicates that the subtask is relatively simple and will be directly processed in a conventional manner, such as using a single model to output the processing result. Conversely, if the second evaluation value is higher than or equal to the second preset value, multiple processing models are called to meet more complex task requirements. The second preset value can be determined based on historical data and dynamically adjusted during the evaluation process. In this way, not only can computing resources be reasonably allocated according to the actual difficulty of the task, but also the efficiency and accuracy of task processing can be significantly improved, ensuring that each subtask can be processed in an optimized manner, thereby overall enhancing the completion quality and success rate of complex tasks.

[0097] In one embodiment, evaluating the processing difficulty of the subtask includes:

[0098] Evaluating the processing difficulty of the subtask based on the task difficulty evaluation criteria, where the task difficulty evaluation criteria include at least one of task complexity, estimated resource consumption, data dependency, depth of domain knowledge, and result uncertainty.

[0099] Optionally, the task difficulty evaluation criteria are used to comprehensively evaluate the difficulty of the generative large model in processing complex tasks, so as to adjust the execution strategy, reasonably allocate resources, and optimize the task process. Through scientific indicators and systematic evaluation methods, this criteria can help the model accurately identify the complexity, high-risk links, and potential execution bottlenecks in the task, thereby significantly improving the efficiency and success rate of task completion.

[0100] The following describes each evaluation indicator in the task difficulty evaluation criteria.

[0101] Optionally, the task complexity indicator is used to evaluate the logical and semantic complexity that needs to be processed by the generative large model in the task. The evaluation content includes and is not limited to whether multi-step answers are required, multi-level semantic understanding, and the complexity of the input content (such as text length, multi-domain integration, etc.).

[0102] Optionally, the estimated resource consumption indicator is used to evaluate the requirements for computing resources (CPU / GPU time), memory occupancy, and processing time when the task is completed, and the score can be obtained by comparing the resource consumption ratio with the system resource upper limit.

[0103] Optionally, the data dependency indicator is used to evaluate the degree of dependence of the task on external data or the output of previous subtasks, and the evaluation method is to analyze the number of data sources required by the task and the data quality requirements for dependence.

[0104] Optionally, the domain knowledge depth metric is used to evaluate the depth requirement of a task for professional knowledge or specific domain background knowledge. The evaluation content includes the number of domains involved in the subtask and the knowledge modules or model capabilities that need to be invoked.

[0105] Optionally, the result uncertainty metric is used to evaluate the uncertainty or risk of the task completion result, including the ambiguity of the task goal and the reliability of external data, and the score can be calculated by analyzing the unclear factors in the subtask and calculating the weights.

[0106] Exemplarily, assume that the subtask is to analyze the background, process, and impact of a specific historical event and provide relevant data references. This task requires a comprehensive and multi-angle discussion of the specific historical event, involving multi-domain knowledge and data references.

[0107] For the task complexity metric, it can include the multi-step ratio and the semantic complexity ratio. The subtask requires the model to generate answers for three main parts (background, process, impact), which is a multi-step task. Assume that the highest complexity is set to 5 steps and the actual number of steps required is 3 steps. The multi-step complexity ratio = 3 / 5 = 0.6. The task involves high-level semantic understanding, which is set to level 4, and the highest level is level 5. The semantic complexity score = 4 / 5 = 0.8. Take the average of the above two scores. Assuming that the weights of the two are equal, the task complexity score = (0.6 + 0.8) / 2 = 1.4 / 2 = 0.7.

[0108] For the resource consumption estimation metric, it can include the input length ratio and the processing time ratio. Assume that the input text length of the task is 1500 words and the maximum processing capacity of the system is 2000 words. Then the input length ratio = 1500 / 2000 = 0.75. It is expected to take 60 seconds, and the maximum processing time allowed by the system is 100 seconds. Then the processing time ratio = 60 / 100 = 0.6. Take the average of the above two scores. Assuming that the weights of the two are equal, the resource consumption estimation score = (0.75 + 0.6) / 2 = 1.35 / 2 = 0.675.

[0109] For the data dependency metric, assume that the current subtask needs to reference 5 external data sources, the maximum number of data sources set by the system is 10, and there is no dependency on previous subtasks. Then the data dependency score = 5 / 10 = 0.5.

[0110] For the domain knowledge depth metric, assume that the subtask involves multiple domains such as history, politics, and economy, and 3 knowledge modules need to be enabled. The maximum number of available knowledge modules in the system is 5. Then the knowledge domain depth score = 3 / 5 = 0.6.

[0111] For the result uncertainty indicator, the subtask description is clear, the goal is definite, the external data source is highly reliable, but there may be some missing data. The estimated uncertainty is 20%, so the result uncertainty score = 20% = 20 / 100 = 0.2.

[0112] Summarize the scores of each indicator. Assuming that the weights of each indicator are equal, the comprehensive difficulty score (i.e., the second evaluation value) is calculated as follows:

[0113] The second evaluation value = (task complexity score + resource consumption score + data dependency score + domain knowledge depth score + result uncertainty score) / 5 = (0.7 + 0.675 + 0.5 + 0.6 + 0.2) / 5 = 0.535.

[0114] Assume that the second preset value is 0.5, that is, the second evaluation value is higher than the second preset value, then this subtask is processed based on multiple deep learning models.

[0115] In one embodiment, before processing the subtask based on multiple deep learning models, the task processing method further includes:

[0116] Evaluate the input data of the subtask based on preset evaluation indicators;

[0117] If the input data of the subtask does not meet the input judgment criteria, adjust the input data of the subtask until it meets the input judgment criteria.

[0118] Optionally, before processing the subtask based on multiple deep learning models, the input data of the subtask can be evaluated according to preset evaluation indicators (such as input judgment criteria). If the input data does not meet the input judgment criteria, adjust the input data until it meets the criteria. Through the pre-evaluation and adjustment mechanism, it is ensured that the input data of each subtask has been optimized before entering the deep learning model for processing, thus significantly improving the accuracy and efficiency of subsequent processing, reducing errors or deviations caused by input data quality problems, and ultimately improving the quality and reliability of the overall task completion. Through the input inspection and adjustment process, the accuracy and integrity of the input data quality are guaranteed, and the reliability and efficiency of task execution are fundamentally improved.

[0119] In one embodiment, processing the subtask based on multiple deep learning models includes:

[0120] Process the subtask based on multiple deep learning models to obtain multiple deep learning model processing results;

[0121] Evaluate the multiple deep learning model processing results respectively based on preset evaluation indicators;

[0122] Fuse the multiple deep learning model processing results based on the evaluation results to obtain the processing result of the subtask.

[0123] Optionally, first, the same subtask is processed in parallel by multiple deep learning models to generate multiple different processing results. Next, each model's output is independently evaluated according to preset evaluation metrics (such as description matching degree, input structure integrity, input accuracy, content consistency, semantic coherence, etc.) to ensure the quality and reliability of each result. Finally, based on these evaluation results, an intelligent fusion algorithm is used to integrate the outputs of multiple models to form the final subtask processing result. In this way, not only can the advantages of different models be fully utilized to improve the accuracy and robustness of the processing result, but also the limitations that may exist in a single model can be effectively avoided through the evaluation and fusion mechanism, thereby significantly improving the efficiency and quality of overall task execution.

[0124] In one embodiment, the preset evaluation metrics include at least one of description matching degree, structural integrity, input accuracy, content consistency, and semantic coherence.

[0125] Optionally, the description matching degree metric is used to evaluate whether the input meets the requirements of the task description, including whether the data content is relevant to the target task. The evaluation method can calculate the matching degree between the input content and the task description using semantic similarity, and determine whether the input covers the key topics or information points required by the subtask.

[0126] Optionally, the structural integrity metric is used to evaluate whether the structure of the input data is complete and whether it contains all the fields or partial content required for task execution. The evaluation methods include but are not limited to checking whether the input contains the necessary fields and formats for each subtask, and determining whether there is missing or redundant information in the input.

[0127] Optionally, the input accuracy metric is used to evaluate whether the content of the input data is correct and there are no obvious logical or factual errors. The evaluation methods include comparing the degree of fit between the input content and the expected rules, knowledge base, or context, and checking whether the input content contains obvious contradictions or errors.

[0128] Optionally, the content consistency metric is used to evaluate whether the input content is logically self-consistent and meets the requirements of the subtask for the dependency relationship of the input data. The evaluation methods include analyzing whether there are internal contradictions or logical conflicts in the input content that do not match the task background, and checking whether the input data conforms to the dependency relationship chain of the subtask.

[0129] Optionally, the semantic coherence metric is used to evaluate whether the expression of the input content is semantically fluent and coherent, and does not cause confusion or semantic jumps in the generation model. The evaluation methods include checking whether there are semantic contradictions or jumps in the input content, and measuring the semantic consistency score of the input paragraph through a natural language model.

[0130] Exemplarily, assume the task is to generate a complete financial report based on the financial data input by the user. The input is required to provide the necessary financial metrics (such as revenue, cost, profit, etc.) and meet the task objectives.

[0131] For the description matching degree metric, calculate the proportion of the number of matching information points to the total number of information points in the task description. The user input provides 5 financial metrics (revenue, cost, profit, tax rate, and cash flow), and the task requires 6 metrics (including the additional return on investment). The description matching degree score = 5 / 6 = 0.833.

[0132] For the structural integrity metric, calculate the proportion of the number of required fields that actually exist to the total number of required fields. The table provided by the user contains 8 fields, of which 6 fields are required for the task, and the remaining 2 are redundant fields. The structural integrity score = 6 / 8 = 0.75.

[0133] For the input accuracy metric, calculate the proportion of the number of correct data points to the total number of data points. Among the 6 required fields, the data in 5 fields is correct, and the data in 1 field (profit) contains a logical error (such as the result of revenue minus cost does not match). The input accuracy score = 5 / 6 = 0.833.

[0134] For the content consistency metric, calculate the proportion of the number of consistent data points to the total number of data points. Among the 5 numerical dependency relationships involved in the input (such as revenue - cost = profit), 4 are logically consistent, and 1 has an error. The content consistency score = 4 / 5 = 0.8.

[0135] For the semantic coherence metric, calculate the proportion of the number of semantically fluent paragraphs to the total number of paragraphs. The input text contains 3 main paragraphs, 2 of which are semantically expressed smoothly, and the other 1 has semantic jumps that make it difficult to understand. The semantic coherence score = 2 / 3 = 0.667.

[0136] Take the average of the scores of each metric, and calculate the comprehensive input quality score = (description matching degree score + structural integrity score + input accuracy score + content consistency score + semantic coherence score) / 5 = (0.833 + 0.75 + 0.833 + 0.8 + 0.667) / 5 = 0.7766.

[0137] In one embodiment, the evaluation result is the evaluation score of the processing results of multiple deep learning models. Based on the evaluation result, fuse the processing results of multiple deep learning models to obtain the processing result of the subtask, including:

[0138] Extract the highest score corresponding to each evaluation metric in the preset evaluation metrics from the evaluation scores of the processing results of multiple deep learning models respectively, as the model processing result corresponding to the evaluation metric;

[0139] Fuse the model processing results corresponding to each evaluation index to obtain the processing result of the subtask.

[0140] Optionally, when processing the subtask, first score the processing results of multiple deep learning models based on preset evaluation indexes to generate the evaluation scores of each model. Subsequently, extract the highest score corresponding to each evaluation index in the preset evaluation indexes respectively as the optimal model processing result. Finally, fuse the model processing results corresponding to these highest scores to form the final subtask processing result. This ensures that the optimal performance under each evaluation index is integrated into the final answer, thus significantly improving the comprehensiveness, accuracy, and consistency of the subtask processing result, and effectively avoiding the biases and limitations that may exist in a single model.

[0141] Optionally, in this embodiment, an answer fusion mechanism based on multiple models is adopted to obtain the processing result of the subtask, ensuring that the fused answer absorbs the advantages of each generated answer, avoiding the biases and limitations of a single answer, and improving the comprehensiveness, accuracy, and consistency of the final output. The evaluation of the answers generated by each model can adopt the same indexes as those for evaluating the input data, or different indexes. Taking the example of adopting the same input judgment criteria, evaluate each generated answer, and respectively obtain the scores of indexes such as description matching degree, structural integrity, input accuracy, content consistency, and semantic coherence. For each index, select the answer content with the highest score from all answers. For example, for the description matching degree: Answer A has the highest score, and its relevant content is selected. For the structural integrity: Answer B has the highest score, and its relevant content is selected. For the input accuracy: Answer C has the highest score, and its relevant content is selected. Then, according to the selected high-score indexes, extract the corresponding content. For example: Extract the part related to the task description matching from Answer A. Extract the part with the best structural integrity from Answer B. Integrate the extracted high-score content in logical order and semantic coherence, avoiding semantic jumps or logical conflicts. Re-check the logical consistency and coherence of the fused content, output the final fused answer, and retain the source of each part of the content and the corresponding high-score evaluation index scores.

[0142] Exemplarily, assume that the subtask is to explain the background, technical implementation, and application impact of a certain technological breakthrough based on a deep learning model. Three answers A, B, and C are generated based on different models. Among them, Answer A explains the background in detail, but there are omissions in the technical implementation part. The technical implementation part of Answer B is detailed, but the semantic expression is not smooth enough. The overall content of Answer C is relatively comprehensive, but the background part lacks details. Evaluate each answer using the input judgment criteria, and obtain the evaluation results shown in Table 1:

[0143] Table 1

[0144]

[0145] Next, select and integrate the high-score content corresponding to each indicator. For the description matching degree, select the relevant content of Answer A (the highest score is 0.8). For the structural integrity, select the relevant content of Answer B (the highest score is 0.9). For the input accuracy, select the relevant content of Answer C (the highest score is 0.9). For the content consistency, select the relevant content of Answer A (the highest score is 0.9). For the semantic coherence, select the relevant content of Answer A (the highest score is 0.8). Then, by integrating the high-score content in the logical order of the task, the final answer may be as follows:

[0146] Background: Based on Answer A, provide detailed background information on technological breakthroughs to ensure matching the task description and logical consistency.

[0147] Technical implementation: Based on Answer B, supplement technical details to make the content structure complete.

[0148] Application impact: Based on Answer C, ensure accuracy and logic, and at the same time expand the application impact.

[0149] Through the above design, the answer fusion mechanism can not only effectively utilize the advantageous parts in multiple answers generated by the generative large model, but also ensure that the finally output answer reaches a high level in all aspects, thereby improving the quality and efficiency of task completion. This systematic fusion method provides strong support for handling complex tasks, ensuring that the generated answers are both comprehensive and accurate.

[0150] The task decomposition method of the embodiments of the present application is applicable to various complex task execution scenarios, including text generation, code generation, and logical reasoning, etc., and has broad practical value. Refer to Figure 2 , which is a specific embodiment of the task processing method provided by the embodiments of the present application. As Figure 2 shown, based on a preset deep learning model, perform a decomposition operation on the target task and evaluate the decomposition result. Determine whether the decomposition evaluation result is lower than the first threshold. If so, decompose the target task again; if not, perform a dynamic difficulty assessment on the decomposed subtasks in turn. Determine whether the difficulty assessment result of the current subtask exceeds the second threshold. If so, perform an optimization inspection process and an input adjustment strategy on the input data of the subtask, then generate multiple answers for the subtask based on multiple model parameters, evaluate and analyze the advantages of each answer and fuse them to obtain the processing result of the subtask. If not, use a conventional single model to process the subtask. After processing is completed, continue to process the next subtask until all the subtasks obtained by decomposing the target task are all processed according to the above process.

[0151] In summary, in the task processing method provided by the above embodiments, the target task is decomposed into multiple subtasks based on a deep learning model; the multiple subtasks are evaluated to determine the first evaluation values of the multiple subtasks; if the first evaluation value is lower than the first preset value, the target task is decomposed again based on the deep learning model until the first evaluation values of the multiple subtasks obtained by decomposing the target task are not lower than the first preset value; if the first evaluation value is not lower than the first preset value, corresponding processing strategies are executed for the subtasks according to the processing difficulties of the subtasks. In this way, by evaluating the task decomposition process to optimize the decomposition of the target task and matching the optimal processing strategies according to the processing difficulties of the decomposed subtasks, the resource utilization efficiency and the automation and intelligence levels of task decomposition are improved.

[0152] Based on the same inventive concept as the foregoing embodiments, an embodiment of the present invention provides a computing device, as Figure 3 shown. The computing device includes: a processor 410 and a memory 411 storing a computer program; wherein, Figure 3 The processor 410 shown is not used to indicate that the number of processors 410 is one, but only to indicate the positional relationship of the processor 410 relative to other devices. In practical applications, the number of processors 410 can be one or more; similarly, Figure 3 the memory 411 shown has the same meaning, that is, it is only used to indicate the positional relationship of the memory 411 relative to other devices. In practical applications, the number of memories 411 can be one or more. When the computer program runs on the processor 410, the above task processing method is implemented.

[0153] The computing device may further include: at least one network interface 412. Each component in the computing device is coupled together through a bus system 413. It can be understood that the bus system 413 is used to realize the connection and communication between these components. In addition to the data bus, the bus system 413 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, in Figure 3 all the various buses are labeled as the bus system 413.

[0154] Among them, the memory 411 can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM, Read Only Memory), a programmable read-only memory (PROM, Programmable Read-Only Memory), an erasable programmable read-only memory (EPROM, Erasable Programmable Read-Only Memory), an electrically erasable programmable read-only memory (EEPROM, Electrically Erasable Programmable Read-Only Memory), a ferromagnetic random access memory (FRAM, ferromagnetic random access memory), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM, Compact Disc Read-Only Memory); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM, Random Access Memory), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as a static random access memory (SRAM, Static Random Access Memory), a synchronous static random access memory (SSRAM, Synchronous Static Random Access Memory), a dynamic random access memory (DRAM, Dynamic Random Access Memory), a synchronous dynamic random access memory (SDRAM, Synchronous Dynamic Random Access Memory), a double data rate synchronous dynamic random access memory (DDR SDRAM, Double Data Rate Synchronous Dynamic Random Access Memory), an enhanced synchronous dynamic random access memory (ESDRAM, Enhanced Synchronous Dynamic Random Access Memory), a sync link dynamic random access memory (SLDRAM, SyncLink Dynamic Random Access Memory), a direct rambus random access memory (DRRAM, Direct Rambus Random Access Memory).The memory 411 described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0155] The memory 411 in the embodiments of the present invention is used to store various types of data to support the operation of the computing device. Examples of such data include: any computer programs for operating on the computing device, such as operating systems and application programs; contact data; phone book data; messages; pictures; videos, etc. Among them, the operating system contains various system programs, such as the framework layer, the core library layer, the driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application programs can include various application programs, such as media players, browsers, etc., for implementing various application services. Here, the program for implementing the method of the embodiments of the present invention can be included in the application programs.

[0156] Based on the same inventive concept as the foregoing embodiments, this embodiment also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. The computer-readable storage medium can be a ferromagnetic random access memory (FRAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory, a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM), etc.; it can also be various devices including one or any combination of the above memories, such as mobile phones, computers, tablet devices, personal digital assistants, etc. When the computer program stored in the computer-readable storage medium is run by a processor, it implements the task processing method applied to the above computing device. For the specific step flow implemented when the computer program is executed by the processor, please refer to Figure 1 the description of the illustrated embodiments, which will not be repeated here.

[0157] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0158] In this document, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion. In addition to the recited elements, other elements not expressly listed may also be included.

[0159] As described above, this is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the recited claims.

Claims

1. A task processing method, characterized in that, Including the following steps: Decompose the target task into multiple subtasks based on a deep learning model; Evaluate the multiple subtasks and determine a first evaluation value of the multiple subtasks; If the first evaluation value is lower than a first preset value, decompose the target task again based on the deep learning model until the first evaluation value of the multiple subtasks obtained by decomposing the target task is not lower than the first preset value; If the first evaluation value is not lower than the first preset value, execute corresponding processing strategies for the subtasks according to the processing difficulty of the subtasks; The executing corresponding task processing strategies for the subtasks according to the processing difficulty of the subtasks includes: Evaluate the processing difficulty of the subtasks and determine a second evaluation value; If the second evaluation value is not lower than a second preset value, process the subtasks based on multiple deep learning models; The processing the subtasks based on multiple deep learning models includes: Process the subtasks based on multiple deep learning models to obtain multiple deep learning model processing results; Evaluate the multiple deep learning model processing results respectively based on a preset evaluation index; Fuse the multiple deep learning model processing results based on the evaluation results to obtain the processing result of the subtask; The evaluation result is the evaluation score of the multiple deep learning model processing results, and the fusing the multiple deep learning model processing results based on the evaluation results to obtain the processing result of the subtask includes: Extract respectively the deep learning model processing results corresponding to the highest scores corresponding to each evaluation index in the preset evaluation index from the evaluation scores of the multiple deep learning model processing results as the model processing results corresponding to the evaluation index; Fuse the model processing results corresponding to each evaluation index to obtain the processing result of the subtask.

2. The method according to claim 1, wherein The evaluating the multiple subtasks includes: Evaluate the multiple subtasks based on a task decomposition evaluation criterion, and the task decomposition evaluation criterion includes at least one of logical coherence, executability, rationality of subtask granularity, coverage integrity, input-output consistency, subtask relevance, and workload balance.

3. The method according to claim 1, wherein The evaluating the processing difficulty of the subtasks includes: Evaluate the processing difficulty of the subtasks based on a task difficulty evaluation criterion, and the task difficulty evaluation criterion includes at least one of task complexity, resource consumption estimation, data dependency, depth of domain knowledge, and result uncertainty.

4. The method according to claim 3, wherein Before the processing the subtasks based on multiple deep learning models, the method further includes: Evaluate the input data of the subtasks based on a preset evaluation index; If the input data of the subtasks does not meet the input judgment criterion, adjust the input data of the subtasks until it meets the input judgment criterion.

5. The method according to claim 1 or 4, characterized in that, The preset evaluation index includes at least one of description matching degree, structural integrity, input accuracy, content consistency, and semantic coherence.

6. A computing device, characterized in that, Including: A processor and a memory for storing executable instructions; wherein, the processor is configured to execute the instructions to implement the task processing method according to any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by a processor, the task processing method according to any one of claims 1 to 5 is implemented.

Images