Self-planning method, device and equipment for mathematical geometry problem of intelligent agent and medium

Abstract

The embodiment of the invention provides a self-planning method for a mathematical geometry problem of an intelligent agent, which can be applied to the technical field of artificial intelligence. The self-planning method for the mathematical geometry problem of the intelligent agent comprises the following steps: generating a seed image according to scene demand information of the mathematical geometry problem; performing complex feature planning on the seed image based on a preset guide search rule until a target mathematical problem set is output; and performing problem solving path matching on at least one target mathematical problem in the target mathematical problem set through a preset language model to generate a problem solving planning path set. The embodiment of the invention further provides a self-planning method and device for the mathematical geometry problem of the intelligent agent, equipment, a storage medium and a program product.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, specifically to the field of intelligent body data processing technology, and more specifically to a self-planning method, device, equipment, medium and product for mathematical geometry problems of an intelligent body. Background Art

[0002] Artificial Intelligence (AI) is an important driving force for the new round of scientific and technological revolution and industrial transformation. It is a new key technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. As an important part of intelligent science, artificial intelligence attempts to understand the essence of intelligence and produce a new intelligent machine (i.e., intelligent agent) that can respond in a similar way to human intelligence.

[0003] Language intelligence can be understood as the ability to "understand" empirical things and "reason" between concepts using natural language-based concepts. As the number of parameters soars, large-scale language models based on Transformer architecture gradually demonstrate their ability to understand and reason about concepts in the form of "Chat".

[0004] The performance of large models can be significantly improved by allowing the language model to gradually participate in the process of decomposing a complex problem into sub-problems step by step and solving them one by one. The intermediate steps in this series of reasoning are called Chain of Thought (CoT). CoT technology can greatly improve the performance of large-scale language models on complex reasoning tasks by enhancing the arithmetic, common sense and reasoning of large models, and well implement the processing of complex tasks, such as self-planning solutions to complex mathematical problems. The self-planning solutions to complex mathematical problems in the existing technology have great limitations in automatic theorem discovery and proof, and the matching efficiency of problem search and problem-solving paths is very low, the consumption of reasoning resources is too large, and it is difficult to achieve further breakthroughs in the processing of mathematical geometry problems. Summary of the invention

[0005] In view of at least one of the technical problems existing in the self-planning solution of complex mathematical problems in the above-mentioned prior art, the embodiments of the present invention aim to provide a self-planning method, device, equipment, medium and product for the mathematical geometry problems of an intelligent agent that can further improve the self-planning efficiency of complex mathematical problems and reduce reasoning resources and costs.

[0006] One aspect of an embodiment of the present invention provides a self-planning method for mathematical geometry problems of an intelligent agent, which includes: generating a seed image according to scene requirement information of the mathematical geometry problem; performing complex feature planning on the seed image based on preset guided search rules until a target mathematical problem set is output; and matching the solution path of at least one target mathematical problem in the target mathematical problem set through a preset language model to generate a solution planning path set.

[0007] According to one embodiment of the present invention, complex feature planning is performed on a seed image based on a preset guided search rule until a target mathematical problem set is output, including: geometric feature construction of the seed image according to the preset guided search rule and preset hardware resource information to generate a first intermediate image set; geometric feature construction of at least one first intermediate image in the first intermediate image set again according to the preset guided search rule and preset hardware resource information to generate a second intermediate image set; and the process of generating the second intermediate image set is repeated repeatedly until each target image in the generated target image set meets a preset deductive condition, and the target image set is output as a target mathematical problem set.

[0008] According to one embodiment of the present invention, complex feature planning is performed on the seed image based on preset guided search rules until the target mathematical problem set is output, which also includes: based on preset priority playback buffer rules, generating symmetrical geometric feature actions that comply with preset symmetry mapping rules in the geometric feature construction process, and the symmetrical geometric feature actions are used to execute the geometric feature construction of complex feature planning.

[0009] According to one embodiment of the present invention, a problem-solving path is matched for at least one target mathematical problem in a target mathematical problem set through a preset language model to generate a problem-solving planning path set, including: executing problem-solving auxiliary feature construction for each target mathematical problem in at least one target mathematical problem according to the preset language model to generate at least one problem-solving planning path to constitute a problem-solving planning path set.

[0010] According to one embodiment of the present invention, after matching the solution path of at least one target mathematical problem in the target mathematical problem set through a preset language model to generate a problem-solving planning path set, it also includes: performing a problem-solving ability evaluation on each problem-solving planning path in the problem-solving planning path set through a preset evaluation model to generate a target problem-solving planning path corresponding to at least one target mathematical problem.

[0011] According to one embodiment of the present invention, in performing a problem-solving ability evaluation on each problem-solving planning path in the problem-solving planning path set through a preset evaluation model to generate a target problem-solving planning path corresponding to at least one target mathematical problem, it includes: obtaining a problem-solving ability evaluation value of each problem-solving planning path in the problem-solving planning path set through the preset evaluation model; completing the problem-solving steps of at least one problem-solving planning path in the problem-solving planning path set according to the problem-solving ability evaluation value and a preset beam search optimization rule, and generating a target problem-solving planning path corresponding to at least one target mathematical problem.

[0012] Another aspect of the embodiments of the present invention provides a method and apparatus for self-planning mathematical geometry problems of an intelligent agent, which includes an image generation module, a problem output module, and a path planning module. The image generation module is used to generate a seed image according to the scene requirement information of the mathematical geometry problem; the problem output module is used to perform complex feature planning on the seed image based on a preset guided search rule until a target mathematical problem set is output; and the path planning module is used to match the solution path of at least one target mathematical problem in the target mathematical problem set through a preset language model to generate a solution planning path set.

[0013] Another aspect of an embodiment of the present invention provides an electronic device, comprising one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the self-planning method for mathematical geometry problems of the above-mentioned intelligent agent.

[0014] Another aspect of an embodiment of the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the self-planning method for mathematical geometry problems of the above-mentioned intelligent agent.

[0015] Another aspect of an embodiment of the present invention provides a computer program product, including a computer program, which, when executed by a processor, implements the self-planning method for mathematical geometry problems of the above-mentioned intelligent agent.

[0016] The self-planning method for solving mathematical geometry problems of an intelligent agent provided by an embodiment of the present invention can at least partially solve the problem of low intelligence level in the process of executing actions of intelligent agents in related technologies, and thus can achieve at least one of the following technical effects: By supporting the proposal and solution of search-guided mathematical geometry problems, a self-planning scheme for efficient mathematical geometry problems based on language models is constructed, which breaks through the low efficiency of existing AI systems in the process of automatic discovery and proof of geometric theorems. At the same computing cost, the planning efficiency is significantly improved compared with traditional planning schemes, and the most extensive geometric theorem library to date is established. In addition, it can achieve efficient processing of ultra-complex data geometry problems that can be assisted in construction, and solve the problem of assisted construction of long-chain mathematical geometry problems. Moreover, an efficient symmetry generation mechanism is realized, which improves the search efficiency of spatial search and the generation efficiency of symmetric geometric problems. Nearly 10 billion mathematical geometric bodies that can be assisted in structure are calculated, including more than 2 / 3 of the geometric bodies that can express geometric symmetry, and can support reasoning planning of geometric problems of various symmetry types. Furthermore, compared with traditional geometric reasoning schemes that rely heavily on high-performance computing clusters, it can achieve efficient operation on ordinary computers, and is easier to achieve low-cost promotion and application on consumer-grade machines, greatly reducing the hardware threshold, and has extremely high commercial application value.

[0017] It should be understood that the above general description and the following detailed description are merely exemplary and illustrative and are not intended to limit the scope of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The above contents and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which: Figure 1 A diagram schematically illustrates an application scenario of a method, device, equipment, medium, and program product for self-planning mathematical geometry problems of an intelligent agent according to an embodiment of the present invention; Figure 2 Schematically showing a flow chart of a method for self-planning a mathematical geometry problem of an intelligent agent according to an embodiment of the present invention; Figure 3A An application scenario diagram schematically illustrates a method for self-planning mathematical geometry problems of an intelligent agent according to an embodiment of the present invention; Figure 3B Another application scenario diagram of the self-planning method for mathematical geometry problems of an intelligent agent according to an embodiment of the present invention is schematically shown; Figure 4 Schematically showing a geometric diagram of a target mathematical problem of a method for self-planning mathematical geometric problems of an intelligent agent according to an embodiment of the present invention; Figure 5 A structural block diagram schematically shows a method and apparatus for self-planning mathematical geometry problems of an intelligent agent according to an embodiment of the present invention; and Figure 6A block diagram of an electronic device suitable for implementing a self-planning method for solving mathematical geometry problems of an intelligent agent according to an embodiment of the present invention is schematically shown.

[0019] The above-mentioned drawings are part of the specification of the embodiments of the present invention, which illustrate exemplary embodiments of the present invention. The attached drawings and the description of the specification are used together to illustrate the principles of the embodiments of the present invention. It should be understood that the above general description of the drawings and the following specific implementations are only exemplary and illustrative, and they cannot limit the scope of the present invention. DETAILED DESCRIPTION

[0020] In order to make the objectives, technical solutions and advantages of the embodiments of the present invention more clearly understood, the spirit of the contents disclosed by the present invention will be clearly explained with the accompanying drawings and detailed descriptions below. After understanding the embodiments of the contents of the present invention, any technician in the relevant technical field can change and modify the techniques taught by the contents of the present invention without departing from the spirit and scope of the contents of the present invention.

[0021] The exemplary embodiments and descriptions of the present invention are used to explain the present invention, but are not intended to limit the present invention. In addition, elements / components with the same or similar reference numerals used in the drawings and embodiments are used to represent the same or similar parts.

[0022] The terms “first”, “second”, etc. used in the present invention do not particularly refer to an order or sequence, nor are they used to limit the present invention. They are only used to distinguish elements or operations described with the same technical terms.

[0023] The directional terms used in the present invention, such as up, down, left, right, front or back, etc., are only used to refer to the directions of the drawings. Therefore, the directional terms used are used to illustrate and not to limit the present invention.

[0024] The words “include,” “including,” “have,” “contain,” etc. used in the present invention are open-ended terms, meaning including but not limited to.

[0025] The term "and / or" used in the present invention includes any or all combinations of the items mentioned.

[0026] Regarding the present invention, "plurality" includes "two" and "more than two"; regarding the present invention, "plurality of groups" includes "two groups" and "more than two groups".

[0027] The terms "substantially" and "approximately" used in the present invention are used to modify any quantity or error that may vary slightly, but these slight changes or errors do not change their essence. Generally speaking, the range of slight changes or errors modified by such terms may be 20% in some embodiments, 10% in some embodiments, 5% in some embodiments, or other values. Those skilled in the art should understand that the aforementioned values ​​can be adjusted according to actual needs and are not limited thereto.

[0028] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0029] In the case of using expressions such as "at least one of A, B, and C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but not be limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or a system having A, B, C, etc.). In the case of using expressions such as "at least one of A, B, or C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, or C" should include but not be limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or a system having A, B, C, etc.). Those skilled in the art should also understand that any transitional conjunctions and / or phrases that substantially represent two or more optional items, whether in the specification, claims, or drawings, should be understood to give the possibility of including one of these items, either of these items, or both of these items. For example, the phrase "A or B" should be understood to include the possibilities of "A" or "B", or "A and B".

[0030] The self-planning solution of complex mathematical problems needs to consider the matching of the formulation and answers (i.e., the solution) of complex mathematical problems, especially for Olympic mathematical problems. Currently, there are few computational methods that can realize the autonomous formulation of problems and strict verification of proposals for Olympic mathematical problems, and there is an even greater lack of self-planning schemes for mathematical geometry problems. Compared with other types of mathematical problems that focus on text, Olympic mathematical geometry problems cannot be planned and adjusted through pre-trained language models because they involve numerical and spatial reasoning. In addition, the controversial axiomatization in Euclidean geometry invalidates the completeness or human readability of interactive proof assistants applicable to other fields. Although specific languages ​​tailored for geometry have been created, their design limitations have prevented the space of Olympic mathematical geometry problems from being fully explored. Specifically, the defects of existing planning schemes for complex mathematical geometry problems are mainly reflected in the following three aspects: (1) Limitations of automatic theorem discovery and proof: Existing AI systems are inefficient in the automated discovery and proof of geometric theorems, and are particularly difficult to handle complex problems that require auxiliary construction (such as adding auxiliary lines, points, circles, etc.).

[0031] (2) Challenges in generating symmetry problems: Olympic geometry problems usually require symmetry (such as mirror symmetry and rotational symmetry), but existing methods lack an efficient symmetry generation mechanism, resulting in inefficient exploration of the problem space.

[0032] (3) Excessive consumption of inference resources: Traditional geometric reasoning systems (such as Alpha Geometry) rely on high-performance computing clusters (such as hundreds of CPU cores + multiple GPUs) and cannot run efficiently on ordinary computers.

[0033] In addition, although there are self-planning solutions for related complex geometric mathematical problems in the prior art (such as TongGeometry), they still cannot achieve a more efficient self-planning effect. Therefore, in view of at least one of the technical problems existing in the self-planning solution of complex mathematical problems in the above-mentioned prior art, the embodiments of the present invention are intended to improve the self-planning efficiency of complex mathematical problems and reduce the reasoning resources and costs of intelligent agents. Self-planning method, device, equipment, medium and product for mathematical geometry problems.

[0034] One aspect of an embodiment of the present invention provides a self-planning method for mathematical geometry problems of an intelligent agent, which includes: generating a seed image according to scene requirement information of the mathematical geometry problem; performing complex feature planning on the seed image based on preset guided search rules until a target mathematical problem set is output; and matching the solution path of at least one target mathematical problem in the target mathematical problem set through a preset language model to generate a solution planning path set.

[0035] Figure 1An application scenario diagram schematically illustrates a method, apparatus, device, medium, and program product for self-planning mathematical geometry problems of an intelligent agent according to an embodiment of the present invention.

[0036] like Figure 1 As shown, the application scenario 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 a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0037] 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 shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0038] 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.

[0039] The server 105 may be a server that provides various services, such as a background management server (only as an example) that provides support for websites browsed by users 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.

[0040] It should be noted that the self-planning method for mathematical geometry problems of the intelligent agent provided in the embodiment of the present invention can generally be executed by the server 105. Accordingly, the self-planning method device for mathematical geometry problems of the intelligent agent provided in the embodiment of the present invention can generally be set in the server 105. The self-planning method for mathematical geometry problems of the intelligent agent provided in the embodiment of the present invention 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 self-planning method device for mathematical geometry problems of the intelligent agent provided in the embodiment of the present invention 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.

[0041] It should be understood that Figure 1The 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 the implementation requirements.

[0042] The following will be based on Figure 1 The scene described by Figure 2~Figure 4 A method for self-planning mathematical geometry problems of an intelligent agent in a disclosed embodiment is described in detail.

[0043] like Figure 2 As shown, one aspect of an embodiment of the present invention provides a self-planning method for mathematical geometry problems of an intelligent agent, which includes operations S201 to S203.

[0044] In operation S201, a seed image is generated according to scene requirement information of a mathematical geometry problem; In operation S202, complex feature planning is performed on the seed image based on a preset guided search rule until a target mathematical problem set is output; and In operation S203, a solution path is matched for at least one target mathematical problem in the target mathematical problem set using a preset language model to generate a solution planning path set.

[0045] The intelligent agent can be the executor of the self-planning method of the embodiment of the present invention, or can be the executor controlled by the self-planning method, and can be a humanoid intelligent robot or other AI device. In addition, the intelligent agent can also be an electronic device with ordinary computer processing capabilities, such as a terminal electronic device such as a computer, PAD and mobile phone.

[0046] The scenario requirement information may be information of a "question-answer matching" scenario of applying mathematical geometry problems according to actual needs of the user, and may be specifically related to the user operating the intelligent agent according to the above application scenario, so that the intelligent agent can provide control information directly related to the execution of the self-planning method according to the operation, which may be embodied in the form of control instructions, codes, or even a combination of graphics and text, and is not specifically limited here. For example, the user performs a touch operation through the terminal touch unit of the intelligent agent device to instruct the intelligent agent device to implement the above self-planning method.

[0047] In addition, the scenario requirement information can also be control information generated by the agent based on the user's actual mathematical geometry problem "question-answer matching" requirements. For example, during the user's dialogue interaction with the agent, the agent autonomously implements the above self-planning method based on the user's dialogue content, thereby significantly improving the user experience. The above scenario requirement information can be bound to or contain other instructions or data related to the execution of the self-planning method.

[0048] The seed image is the starting image data for "problem raising" in the self-planning method of the embodiment of the present invention, and is specifically related to the scene requirements of the mathematical geometry problem. For example, the above-mentioned scene requirement information may include control data related to the self-planning of the above-mentioned mathematical geometry problem, which must be a triangular mathematical geometry problem. In this case, the seed image may be a triangular figure; further, the above-mentioned scene requirement information may also include that the triangle must be an acute triangle, then the triangular figure of the seed image may be an acute triangle. The seed image serves as the starting image data of the above-mentioned self-planning method, and is used as the basis for problem raising. On this basis, the graphic complication processing can be performed to form the required triangular mathematical geometry problem.

[0049] The preset guided search rules may be guiding rules for complex feature planning for the above-mentioned seed image, so that the seed image can realize more complex geometric figures, for example, it may be a guided search rule based on a tree. Among them, complex feature planning is the process of adjusting geometric features according to the geometric shape of the seed image, where the so-called "features" may include geometric features such as midpoints, midlines, vertical lines and circles. The geometric feature adjustment process of complex feature planning may be a process of step-by-step planning according to the preset guided search rules, for example, each step only adds different geometric features to the current geometric figure, so that it can generate different and more complex geometric figures. Therefore, with the help of preset guided search rules, the efficiency of complex feature planning can be greatly improved, and the construction of the target mathematical problem can be completed in the shortest time and in the largest number, while ensuring the accuracy of the target mathematical problem.

[0050] It should be noted that, including the seed image, all subsequent derived and generated more complex geometric figures can record all their geometric features in the form of text, such as the various angles of a triangle, the radius of a circle, the length of a line segment, the midpoint of a line segment, and the relationship between various angles and other regular mathematical content. With the recorded content of this geometric feature, specific mathematical geometry problems can be constructed based on the generated complex geometric figures.

[0051] The target mathematical problem set is a set of target mathematical geometry problems that can be ultimately generated by the self-planning method, wherein the starting figure of each target mathematical problem can be the geometric figure of the seed image, and after at least one complex planning adjustment of the geometric features, the target mathematical problem can ultimately be embodied in the form of super-complex mathematical geometry figures, which can generally meet the topic selection criteria of the Olympic Mathematics Competition (such as multiple solutions, knowledge point coverage, symmetry, etc.), and can be used as an alternative topic for the Olympic Mathematics Competition.

[0052] The preset language model can be a deep learning model that can be trained based on certain text data to realize natural language text understanding and natural language task processing, such as LLM. After pre-training with a large number of mathematical geometry problems of the same type as the target mathematical problem and their solutions and other related data (such as problems-corresponding solution steps-answer results), the preset language model can have the ability to solve mathematical geometry problems of the corresponding type. Therefore, the preset language model can achieve solution matching for the target mathematical problem. However, it should be noted that because the target mathematical problem usually contains geometric graphics content, it is necessary to convert graphics into text. After expressing the mathematical geometry problem in the form of text, the preset language model processes the text information to generate a text-based solution path as the solution result. Therefore, with the help of the language model, the processing efficiency of mathematical geometry problems can be improved, while also avoiding errors that are prone to occur in direct processing of graphics.

[0053] Problem-solving path matching can be understood as the preset language model, when processing the target math problem in text form, will explain the idea of ​​solving the problem or answering the question from the perspective of a normal human being. This can include each step of the problem-solving process and the final result of the problem-solving, and also reflect the expression content of the auxiliary composition (such as auxiliary lines) of the mathematical geometry problem. Among them, the problem-solving path matching can be matched against at least one target math problem in the target math problem set, which can be carried out according to the actual application scenario requirements (such as the requirement is the questions and solution outputs of 10 math geometry problems), or it can be determined according to the actual processing power of the computer (such as the number of cores of the processor of the AI ​​device), so as to significantly improve the matching efficiency.

[0054] The problem-solving planning path set may be a set of problem-solving planning paths, having at least one problem-solving planning path for at least one target mathematical problem. Among them, for each target mathematical problem, there may be at least one problem-solving planning path. The problem-solving planning path may be understood as the path information of the problem-solving steps that occur during the solution process for a certain target mathematical problem, and it also includes the final problem-solving result and the auxiliary composition content required for the corresponding problem-solving. In addition, the problem-solving planning path may also include certain identification information at the same time, which is used to reflect the corresponding relationship with the target mathematical problem and the mapping relationship between its subsequent query retrieval and the storage area, which may be specifically embodied in the form of code.

[0055] Therefore, the embodiment of the present invention realizes the complex feature planning of the seed image by presetting the guided search rules, and can autonomously construct a set of super-complex mathematical geometry problems based on the seed image, and can provide these mathematical geometry problems with matching of the solution paths, that is, by supporting the proposal and solution of search-guided mathematical geometry problems, a self-planning scheme for efficient mathematical geometry problems based on language models is constructed, which breaks through the low efficiency of existing AI systems in the process of automatic discovery and proof of geometric theorems. At the same computing cost, the planning efficiency is significantly improved compared with the traditional planning scheme, and the most extensive geometric theorem library to date is established. In addition, it can realize the efficient processing of super-complex data geometry problems that can be assisted in construction, solve the problem of assisted construction of long-chain mathematical geometry problems, and break through the self-planning problem of active construction and solution of mathematical geometry problems based on language models in the prior art, which has extremely high commercial application value.

[0056] In order to enable those skilled in the art to have a clearer understanding of the self-planning method for the mathematical geometry problem of the above-mentioned intelligent agent in the embodiment of the present invention, the following is further provided: Figure 3A-Figure 4 Description.

[0057] Figure 3A An application scenario diagram of a method for self-planning mathematical geometry problems of an intelligent agent according to an embodiment of the present invention is schematically shown.

[0058] like Figure 2-Figure 3A As shown in the figure, according to an embodiment of the present invention, in operation S202, complex feature planning is performed on the seed image based on a preset guided search rule until the target mathematical problem set is output, including: Performing geometric feature construction on the seed image according to preset guided search rules and preset hardware resource information to generate a first intermediate image set; Re-constructing geometric features of at least one first intermediate image in the first intermediate image set according to a preset guided search rule and preset hardware resource information to generate a second intermediate image set; The process of generating the second intermediate image set is repeated repeatedly until each target image of the generated target image set meets the preset deduction condition, and the target image set is output as the target mathematical problem set.

[0059] like Figure 3A As shown, the lowermost triangle ABC in the figure can be determined as a seed image according to the scene requirement information, wherein the seed image can also be set to a circle, a rectangle, a square, etc. according to the scene requirement information.

[0060] In an embodiment of the present invention, the preset guided search rule may be a preset rule for guiding the seed image to perform geometric feature construction to realize autonomous planning of mathematical problems, which may include a tree-guided search rule based on Markov decision processes (MDP). For example, through a finite state tree modeled by a geometric problem, each state corresponds to a geometric construction step (such as adding points, lines, and circles), during which the search direction of the entire complex feature planning process can be guided by the Markov decision process of the preset guided search rule. Specifically, Figure 3A As shown, the seed image triangle ABC undergoes a geometric construction step to construct the midpoint and perpendicular line of the BC segment, thereby forming the first intermediate image triangle ABC (including the midpoint and perpendicular line).

[0061] The preset hardware resource information is a reflection of the hardware processing capability of the agent, which is specifically reflected in the data parallel processing capability information of the hardware of the agent device itself, such as the number of cores of the central processing unit. For example, each time the geometric feature construction is performed for the previous seed image or other intermediate geometric images, multiple mathematical geometric images can be generated simultaneously according to the number or type of geometric features required (such as center, midpoint, vertical line, circle, etc.) and the data parallel processing capability of the agent itself. Specifically, when the number of cores of the central processing unit is 8, the geometric feature construction of 8 images can be performed in parallel at the same time according to the same seed image.

[0062] Therefore, if Figure 3A As shown, the geometric features of multiple graphics can be constructed in parallel based on the seed image, thereby generating multiple first intermediate images as a first intermediate image set. Further, one or more first intermediate images in the above first intermediate image set can be subjected to the next step of geometric feature construction again based on the above preset guided search rule and preset hardware resource information, generating multiple second intermediate images as a second intermediate image set.

[0063] In this way, geometric feature construction is repeatedly performed on the basis of the previous image. Each time geometric feature construction is performed, a corresponding intermediate image can be generated until at least one more complex target image can be generated. These target images can constitute the above target image set, thereby realizing the tree-guided search process based on the seed image, such as Figure 3A The process from seed image A to intermediate image B, intermediate image C and finally target image D is shown. Figure 4 The triangle ABC shown (midpoint D, midpoint I, perpendicular line HID, perpendicular line bEG, circle O, circle c, circle J and related geometric features) can reflect more mathematical geometric features, and at the same time can also meet the requirements of corresponding mathematical knowledge points (such as a certain geometric theorem), thereby meeting the needs of Olympic mathematical geometry problems.

[0064] It should be noted that the first intermediate image is not limited to being a geometric figure generated by the first geometric feature construction as a seed image. In some embodiments, the first intermediate image can also be an intermediate geometric figure generated by other intermediate images after a certain geometric feature construction. That is, the seed image can also be understood as an intermediate geometric figure generated by other geometric feature construction steps.

[0065] The preset deductive condition can be the end condition of the self-planning process closure that realizes the above-mentioned geometric feature construction, which is preset according to the application scenario requirement information, such as the conditions of equal angles to equal sides, triangles being congruent or similar, etc., determined by the question requirements. In other words, the intermediate geometric figures generated after each geometric feature construction can be traversed and searched based on a specific priority search rule (such as Breadth-First Search) to match the above-mentioned preset deductive conditions for each intermediate geometric figure. When the searched intermediate geometric figure meets the above-mentioned preset deductive conditions (such as the presence of equal angles to equal sides, etc.), the geometric construction process of the intermediate geometric figure is terminated, and then it can be output as a target figure. Each target image in the output target image set can have the ultra-complex geometric figures required by the application scenario, which can better meet the question requirements.

[0066] Therefore, with the help of preset deductive conditions, the search process can be terminated in time according to the application requirements of the scene, thereby significantly improving the search efficiency, avoiding the occurrence of invalid search situations, and greatly reducing redundant calculations. Moreover, with the help of rules such as breadth-first search, the efficiency of search closure can be further improved, thereby further improving the output efficiency of the target mathematical problem set as a whole.

[0067] like Figure 2-3A As shown, according to an embodiment of the present invention, in operation S202, complex feature planning is performed on the seed image based on a preset guided search rule until a target mathematical problem set is output, and further includes: Based on the preset priority playback buffer rule, symmetric geometric feature actions that meet the preset symmetry mapping rules are generated in the geometric feature construction process. The symmetric geometric feature actions are used to execute the geometric feature construction of complex feature planning.

[0068] For most mathematical geometry problems, it is also necessary to consider better symmetry or geometric beauty, such as classical symmetry structures (self-symmetry, mirror symmetry, rotational symmetry, etc.) and classical geometric elements (such as nine-point circle, orthocenter, etc.), but this is also the key technical point that existing AI systems find difficult to break through.

[0069] The preset symmetry mapping rule can be understood as a rule for constructing symmetrical geometric feature actions according to the application scenario requirements and the actual geometric features of the seed figure and the intermediate figure in the above-mentioned geometric feature construction process, so as to ensure that the generated geometric figure has the corresponding symmetrical geometric features after each geometric feature construction. The so-called symmetrical geometric feature action is an operation of performing geometric construction according to symmetrical geometric features, which can be specifically embodied as a certain symmetrical geometric feature, such as a central symmetric feature.

[0070] In the complex feature planning process of the above-mentioned geometric feature construction, a symmetry geometric feature mapping table (i.e., a symmetry feature mapping table) can be constructed based on the search data of the symmetric geometric features, the symmetric geometric features that appear can be pre-stored, and symmetric actions can be automatically generated (such as mirror-symmetrical point and line constructions), so that in the subsequent construction process of symmetric geometric features, the symmetric geometric features of the mapping table can be directly selected to ensure that the graphics generated in each geometric construction process have natural symmetry.

[0071] The preset priority playback buffer rule can be understood as a setting rule for buffering existing symmetrical geometric features by setting a buffer zone to achieve fast query and matching of existing symmetrical geometric features. It can cache high-value intermediate states (such as configurations containing colinear or symmetrical features), thereby further improving the generation efficiency of geometric figures of symmetrical geometric features.

[0072] like Figure 3A As shown, in the process of geometric feature construction of the leftmost seed image, the applied symmetrical geometric features can be pre-stored based on the preset priority buffer rule, and reflected in the symmetrical feature mapping table that conforms to the preset symmetry mapping rule. When the second or other geometric feature construction process is performed, the existing symmetrical geometric features can be searched based on the preset priority buffer rule, thereby avoiding the repetitive search of all geometric features, making the generation of geometric figures with symmetrical geometric features more efficient, and at the same time ensuring that the geometric figures generated by each geometric feature construction have symmetrical features.

[0073] In addition, for geometric feature constructions with symmetric features, such as Figure 3A The midpoint of the line segment AB of the triangle ABC shown is actually the same as the midpoint of the line segment BA. In this way, the symmetry can be used to exclude the situation where the geometric features are actually the same, thereby further improving the efficiency of geometric feature construction and reducing the repetition rate of constructing the same geometric figures.

[0074] Therefore, with the help of the above-mentioned symmetry-guided generation mechanism, the construction efficiency of symmetrical geometric figures can be further improved on the basis of ensuring the generation of symmetrical geometric figures. Therefore, the generation efficiency of the overall target mathematical problem set is further guaranteed on the basis of generation accuracy.

[0075] Figure 3B Another application scenario diagram of the self-planning method for mathematical geometry problems of an intelligent agent according to an embodiment of the present invention is schematically shown.

[0076] like Figure 2-3B As shown, according to an embodiment of the present invention, in operation S203, a solution path matching is performed on at least one target mathematical problem in the target mathematical problem set by using a preset language model to generate a solution planning path set, including: A problem-solving auxiliary feature construction is performed on each target mathematical problem in at least one target mathematical problem according to a preset language model, and at least one problem-solving planning path is generated to form a problem-solving planning path set.

[0077] like Figure 3B As shown, in the problem-solving planning process of the embodiment of the present invention, the processing process of the preset language model 320 for the target mathematical problem can also solve a specified number of target mathematical problems from the target mathematical problem set according to the scenario question-setting requirements. For example, if the question-setting requirement is 10 questions, more than 10 target mathematical problems can be generated accordingly, but only 10 target mathematical problems can be selected from them for the next step of problem-solving planning.

[0078] Before executing the problem-solving planning operation for the selected target math problem, the geometric figures of the target math problem also need to be textualized to ensure that the preset language model can achieve targeted text processing. Among them, the preset language model has undergone large-scale training of a large number of Olympic math problems and problem-solving path matching, has targeted problem-solving capabilities, and can output the problem-solving process in the form of a problem-solving planning path in the form of text. Specifically, all geometric features of mathematical geometric figures in the target math problem, such as triangles, circles, points, medians, perpendicular lines, and symmetry features, can be detected and converted into text form. Figure 4 The target math problem shown is presented in the form of a geometric figure. After being processed by the trained preset language model, the geometric features of the geometric figure can be expressed in text form (part) using the processing action (Action) of the preset language model. Specifically, it is as follows: Action(BaseAcuteTriangle,"","CBA") Action(CircumscribedCircle,"ABC","") Action(AnyArc, "CBO", "P") Action(CircumscribedCircle, "A0P","a") Action(CircumscribedCircle, "BOp", "b") Action(CircumscribedCircle,"C0p","c") Action(PerpendicularLine, "BaC" , "D") Action(PerpendicularLine, "AbC","E") Action(PerpendicularLine, "AcB", "F") Action(IntersectLineLine,"aDbE","I") Action(IntersectLineLine, "aDcF","H") Action(IntersectLineLine,"bEcF","G") Action(CircumscribedCircle,"IHJ","J") … Fact("perp", [Angle(*"JPO")]) Therefore, the preset language model can match the textual content of the above-mentioned target mathematical problems to solve them, thereby improving the efficiency of problem-solving matching while ensuring the accuracy of the solutions, realizing the textual processing of geometric figures, and solving the dilemma that language models are difficult to effectively handle mathematical problems involving geometric figures.

[0079] Regarding the problem-solving planning process, it is very important for solving geometric problems to include the construction of auxiliary features for problem-solving, such as the construction of various auxiliary lines, such as the construction based on points, lines, and circles to assist in the problem-solving process. Based on the large-scale data training of the problem-solving process based on the preset language model, the preset language model can process the geometric auxiliary feature construction in text form when processing the above-mentioned target mathematical problems. Among them, when executing the auxiliary feature construction for each target mathematical problem, detailed problem-solving steps and the auxiliary geometric features used in each problem-solving step can be generated in text form. These problem-solving steps and the problem-solving path information formed by the corresponding geometric features are used to constitute the above-mentioned problem-solving planning path. Specifically, the above-mentioned problem-solving process for example Figure 4 After the solution planning of the target mathematical problem shown in the figure is performed, the text form (part) of the generated solution planning path can be specifically expressed as follows: Action(ExtendEqual,"A0") [1] eqangle ( PA , PBQ ) because … … # Prove BPH collinear [6] para ( BP , HP ) because para ( BQ , FH ) [0],… # Prove APG collinear [7]eqline (AP, GP) because eqline (BP, HP)[6],… # Prove GPQ a right angle [8] perp (GPQ) because eqline ( AP , GP ) [7],… # Angle chase [9] perp ( JP0 ) because perp ( GPQ )[8],… Therefore, based on the language model pre-trained by code in the embodiment of the present invention, possible auxiliary construction actions can be generated, thereby achieving efficient and accurate solutions to target mathematical problems based on text form, while retaining more comprehensive problem-solving path information.

[0080] like Figure 2-3B As shown, according to an embodiment of the present invention, after matching the solution path of at least one target mathematical problem in the target mathematical problem set by a preset language model to generate a solution planning path set in operation S203, the method further includes: A problem-solving ability assessment is performed on each problem-solving planning path in the problem-solving planning path set through a preset assessment model to generate a target problem-solving planning path corresponding to at least one target mathematical problem.

[0081] Considering that the preset language model may generate a problem-solving planning path that cannot completely solve the problem effectively during the process of solving the target math problem, such as the lack of some problem-solving steps. Therefore, the matching relationship between the target math problem and the corresponding problem-solving planning path can be further evaluated to evaluate the problem-solving ability of each problem-solving planning path generated by the preset language model, and the final target problem-solving planning path can be generated based on the intuitive display of the obtained problem-solving ability.

[0082] like Figure 3BAs shown, the preset evaluation model 330 can be a valuation model that has been pre-trained with large-scale data and has a problem-solving ability evaluation function, and can perform problem-solving ability evaluation on each problem-solving planning path provided by the preset language model. The problem-solving ability evaluation process is mainly based on whether the problem-solving planning path provides accurate problem-solving steps and answer results.

[0083] Specifically, the preset evaluation model can predict the problem-solving steps of the target mathematical problem corresponding to the problem-solving planning path, predict the number of problem-solving steps required, and compare it with the problem-solving steps of the actual problem-solving planning path, so as to determine whether there are any missing steps in the actual problem-solving planning path. Then, the problem-solving steps of the problem-solving planning path can be completed according to the missing steps, so as to generate the target problem-solving planning path for each target mathematical problem.

[0084] Therefore, with the help of the preset evaluation model, it is possible to predict the required steps of the problem-solving planning path, and at the same time, it is possible to optimize the reasoning of the problem-solving planning path, so as to cooperate with the preset language model to solve the problem of long-chain auxiliary construction in traditional technology, greatly improving the efficiency and matching accuracy of problem-solving planning.

[0085] like Figure 2-3B As shown, according to an embodiment of the present invention, in performing problem-solving ability evaluation on each problem-solving planning path in the problem-solving planning path set by using a preset evaluation model to generate a target problem-solving planning path corresponding to at least one target mathematical problem, the method includes: Obtaining a problem-solving ability evaluation value of each problem-solving planning path in the problem-solving planning path set through a preset evaluation model; The problem-solving steps of at least one problem-solving planning path in the problem-solving planning path set are completed according to the problem-solving ability evaluation value and the preset beam search optimization rule, so as to generate a target problem-solving planning path corresponding to at least one target mathematical problem.

[0086] like Figure 3B As shown, the preset evaluation model 330 can evaluate the problem-solving ability of each problem-solving planning path generated by the preset language model 320, and output the evaluation value of the problem-solving ability of each problem-solving planning path (i.e., the problem-solving ability evaluation value), which can be presented in the form of numerical values ​​(such as 0-100%) or text (strong, relatively strong, medium, relatively weak, weak), etc., and the evaluation value can be used to achieve accurate feedback on the problem-solving ability of the problem-solving planning path. For example, the evaluation value can correspond to the number of missing problem-solving steps of the problem-solving planning path. For the same problem-solving planning path, the greater the number of missing problem-solving steps, the worse its problem-solving ability is, and the smaller its evaluation value can be.

[0087] Therefore, the problem-solving ability evaluation value can be used to evaluate the problem-solving ability of each problem-solving planning path for the target mathematical problem, so that all problem-solving planning paths can be sorted according to the problem-solving ability evaluation value.

[0088] The preset beam search optimization rule can be a set rule for completing the problem-solving steps for the problem-solving planning path with better problem-solving ability based on the beam search algorithm. For example, the beam search optimization can be used to complete the problem-solving steps for the three problem-solving planning paths with the smallest problem-solving ability evaluation values, thereby generating three target problem-solving planning paths for the target mathematical problem. In this way, the problem-solving efficiency can be greatly improved, while the matching error rate of the problem-solving planning path can be reduced, ensuring the matching accuracy of the problem-solving planning path.

[0089] Furthermore, if Figure 3B As shown, the target problem-solving planning path and its related information generated by the preset evaluation model 330 can be cached by the heap buffer 340, so that when the preset evaluation model 330 performs the next step completion operation of the problem-solving planning path, the preset beam search optimization rule can be used to preferentially compare the cached target problem-solving planning path of the heap buffer 340, thereby reducing unnecessary repeated operations to further improve the problem-solving planning efficiency. Finally, the target problem-solving planning path is outputted by the system reasoning engine 310.

[0090] In summary, the self-planning method for mathematical geometry problems of the intelligent agent of the embodiment of the present invention can guide the formulation and solution of mathematical geometry problems by supporting search, and construct a self-planning scheme for efficient mathematical geometry problems based on language models, breaking through the low efficiency of existing AI systems in the process of automatic discovery and proof of geometric theorems. At the same computing cost, the planning efficiency is significantly improved compared with traditional planning schemes, and the most extensive geometric theorem library to date is established. In addition, it can achieve efficient processing of ultra-complex data geometry problems that can be assisted in construction, and solve the problem of assisted construction of long-chain mathematical geometry problems. Moreover, an efficient symmetry generation mechanism is implemented, which improves the search efficiency of spatial search and the generation efficiency of symmetric geometric problems. Nearly 10 billion mathematical geometric bodies that can be assisted in structure are calculated, including more than 2 / 3 of the geometric bodies that show geometric symmetry, and can support reasoning planning of geometric problems of various symmetry types. Furthermore, compared with traditional geometric reasoning solutions that rely heavily on high-performance computing clusters, it can achieve efficient operation on ordinary computers, and is easier to achieve low-cost promotion and application on consumer-grade machines, which greatly reduces the hardware threshold and has extremely high commercial application value. Specifically, the self-planning method for mathematical geometry problems of the above-mentioned intelligent agent in the embodiment of the present invention is based on data-driven statistical optimization, and uses 196 historical Olympic problems to construct statistical priors to guide the direction of the above-mentioned tree search (such as action type, depth distribution), and improve the relevance of generated problems. And further design automatic evaluation indicators (such as proof trajectory length, symmetry type, etc.) to screen problems that meet the Olympic difficulty and aesthetic standards (geometric symmetry). In addition, with the help of Figure 3B The resource-efficient system reasoning engine shown can further optimize the deductive database rules, integrate algebraic reasoning capabilities, avoid the overhead of independent algebraic modules, and support parallel search (10,368 CPU cores in parallel), traverse 143 million paths in 30 days, generate nearly 10 billion theorems, and finally realize that the system can run on ordinary computers (32-core CPU + single GPU). Therefore, the technical effects of the above self-planning method of the embodiment of the present invention can be specifically reflected in the following contents: (1) Large-scale theorem discovery: Based on the same computational cost as traditional technology, we have generated nearly 10 billion mathematical geometry theorems that can be used for auxiliary construction (more than 2 / 3 of which are symmetric), establishing the largest geometric theorem library to date.

[0091] (2) Olympiad Verification: Among the 10 theorems proposed, one was selected for the 2024 China High School Mathematics League (Beijing Division), and two were shortlisted for the 2024 Ersatz Mathematical Olympiad in the United States.

[0092] (3) Reasoning performance breakthrough: In the IMO-AG-30 benchmark test, it surpassed the human gold medalist (30 / 30 full solution) for the first time, with an average solving time of only about 30 minutes (the traditional technology solution requires 90 minutes + 246-core CPU). In the new data set MO-TG-225, it solved 183 / 225 questions, significantly better than 2 / 225 and 102 / 225 in traditional technology.

[0093] (4) Resource efficiency and accessibility: The system runs on an ordinary computer (32-core CPU + single NVIDIA RTX 4090), which means that all functions can be used on consumer-grade machines, greatly reducing the hardware threshold, making it easier to access, and promoting the popularization of geometric reasoning.

[0094] (5) Expert recognition and generalization ability: The generated solutions (such as IMO 2024 P4) are verified by gold medalists to be consistent with the official answers and can reproduce classic lemmas (such as the nine-point circle theorem). It supports multiple symmetry types (self-symmetry, mirror symmetry, rotational symmetry), covering 86.8% of the representation requirements of IMO plane geometry problems. Based on the above-mentioned self-planning method for mathematical geometry problems of intelligent agents, the present invention also provides a self-planning method device for mathematical geometry problems of intelligent agents. Figure 5 The device is described in detail.

[0095] Figure 5 A structural block diagram schematically shows a method and apparatus for self-planning mathematical geometry problems of an intelligent agent according to an embodiment of the present invention.

[0096] like Figure 5 As shown, the self-planning method device 500 for solving mathematical geometry problems of an intelligent agent in this embodiment includes an image generation module 510 , a problem output module 520 and a path planning module 530 .

[0097] The image generation module 510 is used to generate a seed image according to the scene requirement information of the mathematical geometry problem. In one embodiment, the image generation module 510 can be used to perform the operation S201 described above, which will not be described in detail here.

[0098] The problem output module 520 is used to perform complex feature planning on the seed image based on the preset guided search rule until a target mathematical problem set is output. In one embodiment, the problem output module 520 can be used to perform the operation S202 described above, which will not be described in detail here.

[0099] The path planning module 530 is used to match the solution path of at least one target math problem in the target math problem set through a preset language model to generate a solution planning path set. In one embodiment, the path planning module 530 can be used to perform the operation S203 described above, which will not be repeated here.

[0100] According to an embodiment of the present invention, any multiple modules of the image generation module 510, the question output module 520 and the path planning module 530 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present invention, at least one of the image generation module 510, the question output module 520 and the path planning module 530 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware and firmware or in any appropriate combination of any of them. Alternatively, at least one of the image generation module 510, the question output module 520 and the path planning module 530 can be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding function can be executed.

[0101] Figure 6 A block diagram of an electronic device suitable for implementing a self-planning method for solving mathematical geometry problems of an intelligent agent according to an embodiment of the present invention is schematically shown.

[0102] The electronic device provided by an embodiment of the present invention includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the self-planning method for the mathematical geometry problems of the above-mentioned intelligent agent.

[0103] like Figure 6 As shown, the electronic device 600 according to an embodiment of the present invention includes a processor 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 to a random access memory (RAM) 603. The processor 601 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 601 may also include an onboard memory for caching purposes. The processor 601 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0104] In RAM 603, various programs and data required for the operation of electronic device 600 are stored. Processor 601, ROM 602 and RAM 603 are connected to each other via bus 604. Processor 601 performs various operations of the method flow according to the embodiment of the present invention by executing the program in ROM 602 and / or RAM 603. It should be noted that the program can also be stored in one or more memories other than ROM 602 and RAM 603. Processor 601 can also perform various operations of the method flow according to the embodiment of the present invention by executing the program stored in the one or more memories.

[0105] According to an embodiment of the present invention, the electronic device 600 may further include an input / output (I / O) interface 605, which is also connected to the bus 604. The electronic device 600 may further include one or more of the following components connected to the I / O interface 605: an input portion 606 including a keyboard, a mouse, etc.; an output portion 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 608 including a hard disk, etc.; and a communication portion 609 including a network interface card such as a LAN card, a modem, etc. The communication portion 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage portion 608 as needed.

[0106] The present invention also provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to execute the self-planning method for mathematical geometry problems of the above-mentioned intelligent agent.

[0107] The computer-readable storage medium may be included in the device / apparatus / system described in the above embodiment; or it may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present invention is implemented.

[0108] According to an embodiment of the present invention, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, it may include but is not limited to: 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), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, an apparatus or a device. For example, according to an embodiment of the present invention, the computer-readable storage medium may include the ROM 602 and / or RAM 603 described above and / or one or more memories other than ROM 602 and RAM 603.

[0109] An embodiment of the present invention also includes a computer program product, which includes a computer program, which, when executed by a processor, implements the self-planning method for mathematical geometry problems of the above-mentioned intelligent agent.

[0110] The computer program includes program codes for executing the method shown in the flow chart. When the computer program product is run in a computer system, the program codes are used to enable the computer system to implement the method provided by the embodiment of the present invention.

[0111] The computer program executes the above functions defined in the system / device of the embodiment of the present invention when it is executed by the processor 601. According to the embodiment of the present invention, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0112] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 609, and / or installed from a removable medium 611. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0113] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the processor 601, the above functions defined in the system of the embodiment of the present invention are performed. According to the embodiment of the present invention, the system, device, means, module, unit, etc. described above can be implemented by a computer program module.

[0114] According to an embodiment of the present invention, the program code for executing the computer program provided by the embodiment of the present invention can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (e.g., using an Internet service provider to connect through the Internet).

[0115] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0116] In addition, all actions of acquiring information, signals or data in the present invention are carried out in compliance with the corresponding data protection laws, regulations and policies of the country where they are located, and with the authorization given by the owner of the corresponding device.

[0117] It will be appreciated by those skilled in the art that the features described in the various embodiments and / or claims of the present invention may be combined and / or combined in various ways, even if such combinations and / or combinations are not explicitly described in the present invention. In particular, the features described in the various embodiments and / or claims of the present invention may be combined and / or combined in various ways without departing from the spirit and teachings of the present invention. All of these combinations and / or combinations fall within the scope of the present invention.

[0118] The embodiments of the present invention are described above. However, these embodiments are only for the purpose of illustration, and are not intended to limit the scope of the present invention. Although each embodiment is described above, it does not mean that the measures in each embodiment cannot be used in combination. The scope of the present invention is defined by the attached claims and their equivalents. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present invention.

Claims

1. A self-planning method for mathematical geometry problems of an intelligent agent, characterized in that: include: Generate a seed image according to the scene requirement information of the mathematical geometry problem; Performing complex feature planning on the seed image based on a preset guided search rule until a target mathematical problem set is output; as well as A solution path set is generated by matching at least one target mathematical problem in the target mathematical problem set with a preset language model.

2. The method according to claim 1, characterized in that The step of performing complex feature planning on the seed image based on the preset guided search rule until the target mathematical problem set is output includes: Performing geometric feature construction on the seed image according to preset guided search rules and preset hardware resource information to generate a first intermediate image set; Re-constructing geometric features of at least one first intermediate image in the first intermediate image set according to the preset guided search rule and the preset hardware resource information to generate a second intermediate image set; The process of generating the second intermediate image set is repeated repeatedly until each target image of the generated target image set meets a preset deduction condition, and the target image set is output as the target mathematical problem set.

3. The method according to claim 2, characterized in that The step of performing complex feature planning on the seed image based on the preset guided search rule until the target mathematical problem set is output also includes: Based on the preset priority playback buffer rule, a symmetric geometric feature action that complies with the preset symmetry mapping rule is generated in the geometric feature construction process, and the symmetric geometric feature action is used to execute the geometric feature construction of the complex feature planning.

4. The method according to claim 1, characterized in that: In matching the solution path of at least one target mathematical problem in the target mathematical problem set by using the preset language model to generate a solution planning path set, the method includes: A problem-solving auxiliary feature construction is performed on each target mathematical problem in the at least one target mathematical problem according to a preset language model, and at least one problem-solving planning path is generated to constitute the problem-solving planning path set.

5. The method according to claim 1 or 4, characterized in that: After matching the solution path of at least one target mathematical problem in the target mathematical problem set with the preset language model to generate a solution planning path set, the method further includes: A problem-solving ability assessment is performed on each problem-solving planning path in the problem-solving planning path set through a preset assessment model to generate a target problem-solving planning path corresponding to the at least one target mathematical problem.

6. The method according to claim 5, characterized in that The step of performing problem-solving ability evaluation on each problem-solving planning path in the problem-solving planning path set by using a preset evaluation model to generate a target problem-solving planning path corresponding to the at least one target mathematical problem includes: Obtaining a problem-solving ability evaluation value of each problem-solving planning path in the problem-solving planning path set through the preset evaluation model; The problem-solving steps of at least one problem-solving planning path in the problem-solving planning path set are completed according to the problem-solving ability evaluation value and the preset beam search optimization rule, so as to generate a target problem-solving planning path corresponding to the at least one target mathematical problem.

7. A method and device for self-planning mathematical geometry problems of an intelligent agent, characterized in that: include: An image generation module, used to generate a seed image according to scene requirement information of the mathematical geometry problem; A problem output module, used for performing complex feature planning on the seed image based on a preset guided search rule until a target mathematical problem set is output; as well as The path planning module is used to match the solution path of at least one target mathematical problem in the target mathematical problem set through a preset language model to generate a solution planning path set.

8. An electronic device comprising: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images