Multi-dimensional cognitive state vector simulation agent digital textbook dynamic adaptability evaluation method

Abstract

The application provides a multi-dimensional cognitive state vector simulation agent digital textbook dynamic adaptability evaluation method, and relates to the technical field of artificial intelligence assisted education. The method comprises the following steps: extracting teaching entities and semantic relationships between the entities in a source file of a digital textbook to be evaluated through a multi-modal analysis model, and constructing a knowledge graph; initializing multiple groups of virtual learning agents based on a pre-trained large language model, and configuring differentiated cognitive state vectors for each group of virtual learning agents; driving each group of virtual learning agents to traverse the teaching knowledge graph according to a topological sorting by using a knowledge boundary mask mechanism, obtaining understanding feedback of each node by each group of virtual learning agents, and generating confusion degrees and thought chains in the understanding feedback process; calculating a cognitive block index and an adaptability distribution curve of the source file of the digital textbook by using a weighted summation algorithm, and generating an evaluation report. The application simulates human cognitive processes by using a large language model agent, and evaluates the quality and adaptability of a digital textbook.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence-assisted education technology, and in particular to a dynamic adaptive evaluation method for digital teaching materials using a multi-dimensional cognitive state vector simulation agent. Background Technology

[0002] With the deepening of educational informatization, digital textbooks have become widespread. However, the existing evaluation system for digital textbooks has the following pain points:

[0003] 1. Evaluation lag: It usually relies on user feedback data after publication, and cannot provide early warning of instructional design defects during the editing stage.

[0004] 2. Ignoring individual differences: Traditional expert review is often based on the assumption of an "ideal student," which makes it difficult to simulate the real cognitive obstacles of students with weak foundations or those who are good at certain subjects. As a result, the textbooks are "understandable to experts, but not to students."

[0005] 3. Limited technical means: Existing automated tools are limited to typo correction or plagiarism checking, and cannot understand complex pedagogical logic and the effect of combining text and graphics. Summary of the Invention

[0006] To address the aforementioned technical problems in existing technologies, this invention provides a method and apparatus for dynamic adaptive evaluation of digital teaching materials using multi-dimensional cognitive state vector simulation. The technical solution is as follows:

[0007] On the one hand, a multi-dimensional cognitive state vector simulation agent digital textbook dynamic adaptive evaluation method is provided. This method is implemented by a digital textbook dynamic adaptive evaluation device and includes: S1. Obtain the source files of the digital teaching materials to be evaluated, extract the teaching entities and semantic relationships between entities in the source files of the digital teaching materials through a multimodal parsing model, and construct a directed weighted teaching knowledge graph.

[0008] S2. Initialize multiple sets of virtual learning agents based on the pre-trained large language model, and configure differentiated cognitive state vectors for each set of virtual learning agents; wherein, the cognitive state vectors include the density of the prior knowledge base, the decay coefficient of working memory, the maximum inference depth, and the size of the attention focusing window.

[0009] S3. By driving each group of virtual learning agents to traverse the teaching knowledge graph according to the topological sorting through the knowledge boundary masking mechanism, we can obtain the understanding feedback of each group of virtual learning agents for each node, as well as the confusion level and thought chain in the process of generating understanding feedback.

[0010] S4. Based on the confusion level and thought chain of all virtual learning agents for all nodes, use a weighted statistical algorithm to calculate the cognitive hindrance index and adaptive distribution curve of the digital textbook source file, and generate an evaluation report.

[0011] Optionally, in S1, the teaching entities and semantic relationships between entities in the digital teaching material source files are extracted using a multimodal parsing model to construct a directed weighted instructional knowledge graph, including: By using a multimodal parsing model, text nodes, image nodes, and semantic relationships between text nodes and image nodes in digital teaching material source files are extracted, and a directed weighted teaching knowledge graph is constructed based on text nodes, image nodes, and semantic relationships.

[0012] Optionally, the multiple sets of virtual learning agents in S2 include high-achieving agents and struggling agents.

[0013] The configuration of the learning-optimized agent includes: setting the density of the prior knowledge base to mount the full preceding course vector library, setting the working memory decay coefficient to low forgetting, and setting the maximum inference depth to support multi-hop inference.

[0014] The configuration of the learning-struggling agent includes: setting the density of the prior knowledge base to be mounted with a sparsed pre-course vector library, setting the working memory decay coefficient to high forgetting, limiting the attention focus window size to sliding window mode, and setting the maximum reasoning depth to single-step reasoning.

[0015] Optionally, S3 includes: The virtual learning agent is aided in understanding by using retrieval-enhanced generation techniques, and a knowledge mask matrix is ​​constructed to restrict the virtual learning agent's access to external knowledge.

[0016] For each group of virtual learning agents, obtain the requests from the virtual learning agents to retrieve nodes in the teaching knowledge graph according to the topological sorting, query the existence status of the nodes in the prior knowledge base, if the query result is not null, output the understanding feedback of the node and the confusion level and thought chain in the process of generating the understanding feedback; if the query result is null, force the virtual learning agents to infer the meaning of the nodes only using the context of the current digital teaching materials, if the inference is successful, output the understanding feedback of the node and the confusion level and thought chain in the process of generating the understanding feedback, if the inference fails, trigger an alarm.

[0017] Optionally, S3 also includes: The system determines whether an image node serves as a cognitive scaffold in the process of generating comprehension feedback, and obtains a judgment result indicating whether the image node is a valid or invalid cognitive scaffold.

[0018] Among these, determining whether image nodes act as cognitive scaffolds in the process of generating comprehension feedback includes: The virtual learning agent generates confusion levels and thought processes based on text nodes during the comprehension feedback process.

[0019] The virtual learning agent generates new levels of confusion and thought processes based on text and image nodes during the comprehension feedback process.

[0020] By comparing the level of confusion and the thought chain with the new level of confusion and thought chain, it can be determined whether the image nodes play a cognitive scaffolding role in the process of generating comprehension feedback.

[0021] Optionally, the cognitive impairment index in S4 is shown in equation (1) below: (1) In the formula, Indicates cognitive impairment index, This represents the total number of nodes in the graph. , , Indicates the weighting coefficient. Indicates an indicator function, Indicates the first The perplexity of each node, Indicates the preset threshold. This indicates that the virtual learning agent is at the node. Number of self-correction attempts at the location This indicates the maximum number of self-corrections allowed by default. The cosine similarity represents the feature vectors of the node's graph and text. Represents the text feature vector of a node. This represents the feature vector of the node image.

[0022] Optionally, the method further includes: Nodes whose confusion level exceeds a preset threshold or whose thought chain cannot be closed within the maximum reasoning depth step are marked as cognitive breakpoints. The difference between the thought chains of high-achieving agents and low-achieving agents is calculated, and repair suggestions for cognitive breakpoints are generated based on the difference. The repair suggestions include supplementary definitions, analogy examples, and logical connectors.

[0023] On the other hand, a multi-dimensional cognitive state vector simulation agent digital textbook dynamic adaptive evaluation device is provided. This device is applied to the multi-dimensional cognitive state vector simulation agent digital textbook dynamic adaptive evaluation method. The device includes: The graph construction module is used to obtain the source files of the digital teaching materials to be evaluated, extract the teaching entities and semantic relationships between entities in the source files of the digital teaching materials through a multimodal parsing model, and construct a directed weighted teaching knowledge graph.

[0024] The agent factory module is used to initialize multiple sets of virtual learning agents based on a pre-trained large language model, and to configure a differentiated cognitive state vector for each set of virtual learning agents. The cognitive state vector includes the density of the prior knowledge base, the decay coefficient of working memory, the maximum inference depth, and the size of the attention focusing window.

[0025] The simulation and deduction module is used to drive each group of virtual learning agents to traverse the teaching knowledge graph according to the topological sort by using the knowledge boundary masking mechanism, so as to obtain the understanding feedback of each group of virtual learning agents for each node, as well as the confusion level and thought chain in the process of generating understanding feedback.

[0026] The evaluation calculation module is used to calculate the cognitive hindrance index and adaptive distribution curve of the digital textbook source file based on the confusion level and thought chain of all virtual learning agents for all nodes, using a weighted statistical algorithm, and generate an evaluation report.

[0027] Optionally, the map building module is further used for: By using a multimodal parsing model, text nodes, image nodes, and semantic relationships between text nodes and image nodes in digital teaching material source files are extracted, and a directed weighted teaching knowledge graph is constructed based on text nodes, image nodes, and semantic relationships.

[0028] Optionally, multiple sets of virtual learning agents include high-achieving agents and struggling learners.

[0029] The configuration of the learning-optimized agent includes: setting the density of the prior knowledge base to mount the full preceding course vector library, setting the working memory decay coefficient to low forgetting, and setting the maximum inference depth to support multi-hop inference.

[0030] The configuration of the learning-struggling agent includes: setting the density of the prior knowledge base to be mounted with a sparsed pre-course vector library, setting the working memory decay coefficient to high forgetting, limiting the attention focus window size to sliding window mode, and setting the maximum reasoning depth to single-step reasoning.

[0031] Optionally, the simulation module is further used for: The virtual learning agent is aided in understanding by using retrieval-enhanced generation techniques, and a knowledge mask matrix is ​​constructed to restrict the virtual learning agent's access to external knowledge.

[0032] For each group of virtual learning agents, obtain the requests from the virtual learning agents to retrieve nodes in the teaching knowledge graph according to the topological sorting, query the existence status of the nodes in the prior knowledge base, if the query result is not null, output the understanding feedback of the node and the confusion level and thought chain in the process of generating the understanding feedback; if the query result is null, force the virtual learning agents to infer the meaning of the nodes only using the context of the current digital teaching materials, if the inference is successful, output the understanding feedback of the node and the confusion level and thought chain in the process of generating the understanding feedback, if the inference fails, trigger an alarm.

[0033] Optionally, the simulation module is further used for: The system determines whether an image node serves as a cognitive scaffold in the process of generating comprehension feedback, and obtains a judgment result indicating whether the image node is a valid or invalid cognitive scaffold.

[0034] Among these, determining whether image nodes act as cognitive scaffolds in the process of generating comprehension feedback includes: The virtual learning agent generates confusion levels and thought processes based on text nodes during the comprehension feedback process.

[0035] The virtual learning agent generates new levels of confusion and thought processes based on text and image nodes during the comprehension feedback process.

[0036] By comparing the level of confusion and the thought chain with the new level of confusion and thought chain, it can be determined whether the image nodes play a cognitive scaffolding role in the process of generating comprehension feedback.

[0037] Optionally, the cognitive impairment index is shown in equation (1) below: (1) In the formula, Indicates cognitive impairment index, This represents the total number of nodes in the graph. , , Indicates the weighting coefficient. Indicates an indicator function, Indicates the first The perplexity of each node, Indicates the preset threshold. This indicates that the virtual learning agent is at the node. Number of self-correction attempts at the location This indicates the maximum number of self-corrections allowed by default. The cosine similarity represents the feature vectors of the node's graph and text. Represents the text feature vector of a node. This represents the feature vector of the node image.

[0038] Optionally, it also includes: Nodes whose confusion level exceeds a preset threshold or whose thought chain cannot be closed within the maximum reasoning depth step are marked as cognitive breakpoints. The difference between the thought chains of high-achieving agents and low-achieving agents is calculated, and repair suggestions for cognitive breakpoints are generated based on the difference. The repair suggestions include supplementary definitions, analogy examples, and logical connectors.

[0039] On the other hand, a dynamic adaptive evaluation device for digital teaching materials is provided. The device includes a processor and a memory, wherein computer-readable instructions are stored in the memory. When the computer-readable instructions are executed by the processor, any one of the methods described above in the multidimensional cognitive state vector simulation agent digital teaching material dynamic adaptive evaluation method is implemented.

[0040] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement any of the above-described methods of dynamic adaptive evaluation of digital teaching materials based on multidimensional cognitive state vector simulation.

[0041] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: This invention constructs a multimodal teaching knowledge graph containing text, images, and logical relationships, and instantiates multiple virtual learning agents with differentiated cognitive state vectors using a large language model. The cognitive state vectors include parameters such as prior knowledge density, working memory decay rate, and logical reasoning depth. The system drives the agents to perform traversal simulation learning within the knowledge graph, simulating the knowledge retrieval limitations of students with different learning levels through knowledge boundary masking technology, and calculating the cognitive hindrance index and multimodal semantic consistency score of the teaching materials based on a thought chain break detection algorithm. This invention solves the technical problems of traditional textbook evaluation relying on high costs, strong subjectivity, and the inability to quantify cognitive load, achieving digital pre-visualization and accurate assessment of the appropriateness of textbook content. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart of a dynamic adaptive evaluation method for a multi-dimensional cognitive state vector simulation agent digital teaching material provided by an embodiment of the present invention; Figure 2This is an architecture diagram of a multi-dimensional cognitive state vector simulation agent digital textbook dynamic adaptive evaluation method provided by an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the control principle of the cognitive state vector on the agent capability provided in an embodiment of the present invention; Figure 4 This is an implementation effect diagram of the image-text semantic complementarity evaluation provided in the embodiments of the present invention; Figure 5 This is a flowchart of the cognitive impairment index calculation logic provided in the embodiments of the present invention; Figure 6 This is a block diagram of a multidimensional cognitive state vector simulation agent digital textbook dynamic adaptive evaluation device provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of a digital teaching material dynamic adaptability evaluation device provided in an embodiment of the present invention. Detailed Implementation

[0044] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0045] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0046] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0047] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0048] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0049] This invention provides a method for dynamic adaptive evaluation of digital teaching materials using a multi-dimensional cognitive state vector simulation agent. This method can be implemented using a dynamic adaptive evaluation device for digital teaching materials, which can be a terminal or a server. Figure 1 The flowchart shown is for a dynamic adaptive evaluation method of a multidimensional cognitive state vector simulation agent for digital teaching materials. The processing flow of this method may include the following steps:

[0050] S1. Obtain the source files of the digital teaching materials to be evaluated, extract the teaching entities and semantic relationships between entities in the source files of the digital teaching materials through a multimodal parsing model, and construct a directed weighted teaching knowledge graph.

[0051] Optionally, the step S1 above, which involves extracting teaching entities and semantic relationships between entities from the digital teaching material source file using a multimodal parsing model to construct a directed weighted instructional knowledge graph, specifically includes: By using a multimodal parsing model, text nodes, image nodes, and semantic relationships between text nodes and image nodes in digital teaching material source files are extracted, and a directed weighted teaching knowledge graph is constructed based on text nodes, image nodes, and semantic relationships.

[0052] In one feasible implementation, the data is first graphed to construct a directed weighted instructional knowledge graph. ,in For knowledge nodes, These are logically related edges.

[0053] S2. Initialize multiple sets of virtual learning agents based on the pre-trained large language model, and configure differentiated cognitive state vectors for each set of virtual learning agents; wherein, the cognitive state vectors include the density of the prior knowledge base, the decay coefficient of working memory, the maximum inference depth, and the size of the attention focusing window.

[0054] Optionally, the multiple sets of virtual learning agents in S2 may include high-achieving agents and struggling agents.

[0055] The configuration of the learning-optimized agent includes: setting the density of the prior knowledge base to mount the full preceding course vector library, setting the working memory decay coefficient to low forgetting, and setting the maximum inference depth to support multi-hop inference.

[0056] The configuration of the learning-struggling agent includes: setting the density of the prior knowledge base to be mounted with a sparsed pre-course vector library, setting the working memory decay coefficient to high forgetting, limiting the attention focus window size to sliding window mode, and setting the maximum reasoning depth to single-step reasoning.

[0057] In one feasible implementation, agent instantiation is performed: based on a pre-trained large language model, at least two sets of virtual learning agents are initialized, and a differentiated CSV (Cognitive State Vector) is configured for each set of agents. ;in, To determine the density of the prior knowledge base, The working memory decay factor, For maximum reasoning depth, Set the size of the window to focus attention.

[0058] The virtual learning agent may include high-achieving agents and struggling agents, and may also include other types of agents, which are not limited here.

[0059] Specifically, the optimal agent configuration: Mount the full preceding course vector library. Set to 0.05-0.1 (low forgetting). (Supports multi-hop reasoning); Assistance configuration for struggling learners: Mount a sparsed vector library and randomly discard 30%-50% of the non-core concept vectors; Set it to 0.4-0.6 (high forgetting). It is restricted to a sliding window mode, retaining only the most recent N tokens.

[0060] S3. By driving each group of virtual learning agents to traverse the teaching knowledge graph according to the topological sorting through the knowledge boundary masking mechanism, we can obtain the understanding feedback of each group of virtual learning agents for each node, as well as the confusion level and thought chain in the process of generating understanding feedback.

[0061] Optionally, step S3 above may include: The virtual learning agent is aided in understanding by using retrieval-enhanced generation techniques, and a knowledge mask matrix is ​​constructed to restrict the virtual learning agent's access to external knowledge.

[0062] For each group of virtual learning agents, obtain the requests from the virtual learning agents to retrieve nodes in the teaching knowledge graph according to the topological sorting, query the existence status of the nodes in the prior knowledge base, if the query result is not null, output the understanding feedback of the node and the confusion level and thought chain in the process of generating the understanding feedback; if the query result is null, force the virtual learning agents to infer the meaning of the nodes only using the context of the current digital teaching materials, if the inference is successful, output the understanding feedback of the node and the confusion level and thought chain in the process of generating the understanding feedback, if the inference fails, trigger an alarm.

[0063] In one feasible implementation, a constrained simulation is performed: the virtual learning agent is driven to traverse the graph according to topological sorting. During the traversal, a knowledge boundary masking mechanism is used to restrict the agent's access to external knowledge, allowing the agent to retrieve only knowledge based on the current context window and permitted access. Generate node-based feedback.

[0064] Specifically, the knowledge boundary masking mechanism is implemented as follows: When using RAG (Retrieval-augmented Generation) technology to assist agent understanding, a knowledge mask matrix is ​​constructed. When the proxy requests to retrieve the concept When querying the concept, The existence state in: (1) like Returns an empty value, forcing the agent to infer meaning only from the current textbook context; if the inference fails, it triggers a "concept undefined" alarm.

[0065] Furthermore, cognitive breakpoint detection: Real-time monitoring of the agent's perplexity and mental chain integrity during the generation of comprehension feedback; when the perplexity exceeds a preset threshold... Or thought chain If a loop cannot be closed within a step, mark the current node as a cognitive breakpoint.

[0066] Optionally, step S3 above further includes: The system determines whether an image node serves as a cognitive scaffold in the process of generating comprehension feedback, and obtains a judgment result indicating whether the image node is a valid or invalid cognitive scaffold.

[0067] Among these, determining whether image nodes act as cognitive scaffolds in the process of generating comprehension feedback includes: The virtual learning agent generates confusion levels and thought processes based on text nodes during the comprehension feedback process.

[0068] The virtual learning agent generates new levels of confusion and thought processes based on text and image nodes during the comprehension feedback process.

[0069] By comparing the level of confusion and the thought chain with the new level of confusion and thought chain, it can be determined whether the image nodes play a cognitive scaffolding role in the process of generating comprehension feedback.

[0070] S4. Based on the confusion level and thought chain of all virtual learning agents for all nodes, use a weighted statistical algorithm to calculate the cognitive hindrance index and adaptive distribution curve of the digital textbook source file, and generate an evaluation report.

[0071] In one feasible implementation, the evaluation index is calculated as follows: based on the full graph traversal results, a weighted statistical algorithm is used to calculate the cognitive blocking index (CBI) and the adaptive distribution curve of the teaching materials, and an evaluation report is generated.

[0072] Optionally, the formula for calculating the cognitive impairment index in S4 is: (2) In the formula, Indicates cognitive impairment index, This represents the total number of nodes in the graph. , , Indicates the weighting coefficient. Indicates the indicator function, when the perplexity When the threshold is exceeded, the value is 1. Indicates the first The perplexity of each node, Indicates the preset threshold. This indicates that the virtual learning agent is at the node. Number of self-correction attempts at the location This indicates the maximum number of self-corrections allowed by default. The cosine similarity represents the feature vectors of the node's graph and text. Represents the text feature vector of a node. This represents the feature vector of the node image.

[0073] Optionally, the method further includes: Nodes whose confusion level exceeds a preset threshold or whose thought chain cannot be closed within the maximum reasoning depth step are marked as cognitive breakpoints. The difference between the thought chains of high-achieving agents and low-achieving agents is calculated, and repair suggestions for cognitive breakpoints are generated based on the difference. The repair suggestions include supplementary definitions, analogy examples, and logical connectors.

[0074] In one feasible implementation, the dynamic repair suggestion generation step is as follows: when a cognitive breakpoint is detected, based on the difference set between the successful reasoning path of the high-achieving agent and the failed path of the low-achieving agent, "cognitive scaffolding" content is automatically generated. The scaffolding content includes supplementary definitions, analogical examples, or logical connectors.

[0075] This invention views the LLM (Large Language Model) as a cognitively capable "brain," and by injecting different cognitive state vectors, it "reduces" or "specializes" it into virtual students (Agents) with specific knowledge deficiencies, memory limitations, and reasoning habits. These Agents are then "test-run" within the knowledge graph constructed from the teaching materials, and the quality of the teaching materials is evaluated by monitoring their runtime "lag."

[0076] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment will demonstrate how to use this method to evaluate the quality of instructional design for the chapter on "Where Does Water Go (Evaporation)" in elementary school science textbooks. The overall system architecture diagram is as follows. Figure 2 As shown.

[0077] Step S1: Construction of the multimodal graph of the teaching materials: The system receives PDF files from elementary school science textbooks and parses the content of the target chapter. Text node (Core definition paragraph): "Wet clothes dry in the sun. The puddles on the ground also disappear. This liquid water turns into invisible gaseous water vapor and evaporates into the air. This phenomenon is called evaporation."

[0078] Image nodes (Accompanying illustration): A still photograph showing cracked earth under sunlight (Note: This picture only shows the result of "drying", not the process of water turning into gas).

[0079] Semantic analysis: The system identifies For third-grade elementary school students, this is a set of highly abstract key terms. .

[0080] Map Construction: Establishing and The multimodal associated edges are identified, and the node depends on the prior "morphology of water" basic knowledge.

[0081] Step S2: Constructing a differentiated cognitive agent (configuration for elementary school scenarios), such as... Figure 3 As shown: The system instantiates two sets of comparison agents and simulates the cognitive characteristics of primary school students using CSV parameters: Agent A (High-achieving elementary school student control group): Prior knowledge base (Dense): Loads a relatively complete common knowledge base for lower elementary grades, including a rudimentary understanding of basic scientific vocabulary.

[0082] Working memory (Longer): The memory window is set to 2048 Tokens to retain the introductory content from the beginning of this lesson.

[0083] Depth of Reasoning (Medium): Allows for simple textual logical reasoning (such as "because A, therefore B").

[0084] Agent B (Experimental Group of Elementary School Students with Concrete Thinking Dependence / Learning Difficulties): Prior knowledge base (Extremely sparse): Perform a high-proportion random masking operation on the common sense base, retaining only the most basic everyday words (such as "water", "sun", "dry"), and masking the vector indexes of scientific terms such as "gas" and "water vapor".

[0085] Working memory (Extremely short sliding window): The memory window is truncated to 512 tokens (about one page of textbook content), simulating the characteristics of primary school students who are easily distracted and quickly forget the preparatory material.

[0086] Depth of Reasoning (Shallow Level): Restricted to single-step reasoning. The system prompt injects a mandatory instruction: "You are a primary school student who must see a concrete image to understand an abstract process. If the text says 'becomes something invisible' and the image does not show this change, you must report 'cannot imagine' and show extreme confusion."

[0087] Step S3: Simulation loop and cognitive breakpoint detection, such as Figure 4 As shown: The system feeds text nodes to two agents. and image nodes And perform forced RAG cycle monitoring.

[0088] Scenario 1: Text Understanding Monitoring Agent A searches for "water vapor" in its own knowledge base and obtains a vague definition of "a gas that water turns into," which, although abstract, is barely acceptable. The generated CoT (Chain of Thought) shows smooth comprehension, while PPL (Perplexity) is at a low to medium level.

[0089] Agent B returned null values ​​when searching for "water vapor" and "gaseous state" due to knowledge base masking. The system labeled these words as "[unfamiliar concepts]" in the input. Faced with the text "transformed into water vapor, which we cannot see," LLM was severely confused.

[0090] Agent B's CoT monitoring log: "Water is visible, how can it become invisible? What does 'gas' look like? How do I know it's gone into the air? I don't understand this sentence."

[0091] PPL monitoring results: Agent B's PPL value spiked to 95 when generating this interpretation (far exceeding the preset perplexity threshold). The system has marked this point as a critical cognitive breakpoint.

[0092] Scenario 2: Multimodal Consistency Monitoring The system further evaluates the image. Does the photo of cracked soil serve as a cognitive scaffold?

[0093] Agent B combined with image Trying to understand again. The CoT update reads: "I looked at the photo, and the ground was dry. This only tells me the result is that the ground was dry, but it still doesn't tell me how the water 'ran' into the air. The picture doesn't depict the process of the water 'transforming' and 'escaping'."

[0094] Consistency calculation: The confusion level of agent B when reading "text + image" is almost the same as when reading "text only".

[0095] Conclusion: Image It is an ineffective cognitive scaffold. For elementary school students who rely on concrete thinking, it only presents static results and fails to help them bridge the process-oriented cognitive gap from "visible liquid" to "invisible gas".

[0096] Step S4: Indicator calculation and report generation, such as Figure 5 As shown: CBI Index Calculation: Due to the severe mental chain break and high PPL alarm in agent B's understanding of core concepts, and the ineffective combination of text and graphics, according to Formula 2, the CBI score for this chapter for struggling learners is extremely high (e.g., 0.85 out of 1.0), indicating a serious risk of cognitive impairment.

[0097] Automated revision suggestion generation: The system automatically generates evaluation reports and recommendations based on the failure mode analysis of agent B.

[0098] Evaluation Conclusion: The "Evaporation" section of Chapter X in the third-grade science textbook is not well-suited for students who rely heavily on concrete thinking. The abstract process description in the text (transforming into an invisible gaseous state) lacks effective visual support; the current illustration only shows the result and is therefore invalid.

[0099] Revision suggestion: It is strongly recommended to replace static illustrations. Replace it with a dynamic diagram or comic strip that can visually demonstrate the "change process". For example, draw a wavy line with arrows rising from the surface of a puddle, and add cartoonish text "The water turned into water vapor and went up into the sky", in line with children's cognitive principle of "seeing is believing".

[0100] This embodiment powerfully demonstrates that the present invention can realistically simulate the cognitive limitations of primary school students by adjusting the cognitive state vector parameters, accurately identify the breakpoints in the teaching materials that adults take for granted but which pose significant learning obstacles for children, and provide an objective and scientific basis for the compilation of teaching materials.

[0101] In this embodiment of the invention, a multimodal teaching knowledge graph containing text, images, and logical relationships is constructed, and multiple virtual learning agents with differentiated cognitive state vectors are instantiated using a large language model. The cognitive state vectors include parameters such as prior knowledge density, working memory decay rate, and logical reasoning depth. The system drives the agents to perform traversal simulation learning within the knowledge graph, simulating the knowledge retrieval limitations of students with different learning levels through knowledge boundary masking technology, and calculating the cognitive hindrance index and multimodal semantic consistency score of the teaching materials based on a thought chain break detection algorithm. This invention solves the technical problems of traditional textbook evaluation relying on high costs, strong subjectivity, and inability to quantify cognitive load, achieving digital pre-visualization and accurate assessment of the appropriateness of textbook content.

[0102] Figure 6 This is a block diagram illustrating a dynamic adaptive evaluation device for a multidimensional cognitive state vector simulation agent digital teaching material, according to an exemplary embodiment. The device is used in a multidimensional cognitive state vector simulation agent digital teaching material dynamic adaptive evaluation method. (Refer to...) Figure 6 The device includes a map construction module 310, a proxy factory module 320, a simulation and deduction module 330, and an evaluation and calculation module 340. Among them:

[0103] The graph construction module 310 is used to obtain the source files of the digital teaching materials to be evaluated, extract the teaching entities and semantic relationships between entities in the source files of the digital teaching materials through a multimodal parsing model, and construct a directed weighted teaching knowledge graph.

[0104] Specifically, the graph construction module 310 is configured to execute OCR, ASR, and NLP algorithms to convert unstructured teaching materials into structured graphs.

[0105] The agent factory module 320 is used to initialize multiple sets of virtual learning agents based on a pre-trained large language model, and to configure a differentiated cognitive state vector for each set of virtual learning agents. The cognitive state vector includes the density of the prior knowledge base, the decay coefficient of working memory, the maximum inference depth, and the size of the attention focusing window.

[0106] Specifically, the agent factory module 320 is configured to manage the cognitive state vector configuration table and instantiate different types of Agent objects.

[0107] The simulation and deduction module 330 is used to drive each group of virtual learning agents to traverse the teaching knowledge graph according to the topological sort by using the knowledge boundary masking mechanism, so as to obtain the understanding feedback of each group of virtual learning agents to each node, as well as the confusion level and thought chain in the process of generating understanding feedback.

[0108] Specifically, the simulation engine 330 is configured to provide a sandbox environment, execute the agent's reading, retrieval, and inference loops, and record a full set of interaction logs.

[0109] The evaluation calculation module 340 is used to calculate the cognitive resistance index and adaptive distribution curve of the digital textbook source file based on the confusion and thought chain of all virtual learning agents for all nodes, using a weighted statistical algorithm, and generate an evaluation report.

[0110] Specifically, the evaluation calculation module 340 is configured to perform the calculation of Formula 2 and output a heat map visualization report.

[0111] In this embodiment of the invention, a multimodal teaching knowledge graph containing text, images, and logical relationships is constructed, and multiple virtual learning agents with differentiated cognitive state vectors are instantiated using a large language model. The cognitive state vectors include parameters such as prior knowledge density, working memory decay rate, and logical reasoning depth. The system drives the agents to perform traversal simulation learning within the knowledge graph, simulating the knowledge retrieval limitations of students with different learning levels through knowledge boundary masking technology, and calculating the cognitive hindrance index and multimodal semantic consistency score of the teaching materials based on a thought chain break detection algorithm. This invention solves the technical problems of traditional textbook evaluation relying on high costs, strong subjectivity, and inability to quantify cognitive load, achieving digital pre-visualization and accurate assessment of the appropriateness of textbook content.

[0112] Figure 7 This is a schematic diagram of the structure of a digital textbook dynamic adaptability evaluation device provided in an embodiment of the present invention, as shown below. Figure 7 As shown, the digital textbook dynamic adaptability evaluation device may include the above-mentioned Figure 6 The multidimensional cognitive state vector simulation agent digital textbook dynamic adaptive evaluation device is shown. Optionally, the digital textbook dynamic adaptive evaluation device 410 may include a first processor 2001.

[0113] Optionally, the digital textbook dynamic adaptability evaluation device 410 may also include a memory 2002 and a transceiver 2003.

[0114] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0115] The following is combined with Figure 7 A detailed introduction to each component of the digital textbook dynamic adaptability evaluation device 410: The first processor 2001 is the control center of the digital textbook dynamic adaptability evaluation device 410. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0116] Optionally, the first processor 2001 can perform various functions of the digital teaching material dynamic adaptive evaluation device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0117] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 7 CPU0 and CPU1 are shown in the diagram.

[0118] In a specific implementation, as one example, the digital textbook dynamic adaptability evaluation device 410 may also include multiple processors, for example... Figure 7 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0119] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0120] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be accessed through the interface circuit of the digital teaching material dynamic adaptability evaluation device 410. Figure 7 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0121] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0122] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 7 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0123] Optionally, the transceiver 2003 can be integrated with the first processor 2001, or it can exist independently and be connected to the interface circuit of the digital teaching material dynamic adaptive evaluation device 410. Figure 7 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0124] It should be noted that, Figure 7 The structure of the digital teaching material dynamic adaptive evaluation device 410 shown in the figure does not constitute a limitation on the router. The actual knowledge structure identification device may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0125] Furthermore, the technical effect of the digital textbook dynamic adaptability evaluation device 410 can be referred to the technical effect of the multi-dimensional cognitive state vector simulation agent digital textbook dynamic adaptability evaluation method described in the above method embodiments, and will not be repeated here.

[0126] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0127] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0128] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0129] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0130] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0131] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0132] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0133] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0134] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0135] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0136] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0137] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0138] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for evaluating the dynamic adaptability of digital teaching materials using a multi-dimensional cognitive state vector simulation agent, characterized in that: The method includes: S1. Obtain the source files of the digital teaching materials to be evaluated, extract the teaching entities and semantic relationships between entities in the source files of the digital teaching materials through a multimodal parsing model, and construct a directed weighted teaching knowledge graph; S2. Initialize multiple sets of virtual learning agents based on the pre-trained large language model, and configure differentiated cognitive state vectors for each set of virtual learning agents; wherein, the cognitive state vectors include the density of the prior knowledge base, the decay coefficient of working memory, the maximum inference depth, and the size of the attention focusing window; S3. By driving each group of virtual learning agents to traverse the teaching knowledge graph according to the topological sorting through the knowledge boundary masking mechanism, we can obtain the understanding feedback of each group of virtual learning agents for each node, as well as the confusion level and thought chain in the process of generating understanding feedback. S4. Based on the confusion level and thought chain of all virtual learning agents for all nodes, use a weighted statistical algorithm to calculate the cognitive hindrance index and adaptive distribution curve of the digital textbook source file, and generate an evaluation report.

2. The method for dynamic adaptive evaluation of multidimensional cognitive state vector simulation agent digital teaching materials according to claim 1, characterized in that, The step S1 involves extracting teaching entities and semantic relationships between entities from digital textbook source files using a multimodal parsing model, and constructing a directed weighted instructional knowledge graph, including: By using a multimodal parsing model, text nodes, image nodes, and semantic relationships between text nodes and image nodes in digital teaching material source files are extracted, and a directed weighted teaching knowledge graph is constructed based on text nodes, image nodes, and semantic relationships.

3. The method for dynamic adaptive evaluation of digital teaching materials using multidimensional cognitive state vector simulation agents according to claim 1, characterized in that, The multiple virtual learning agents in S2 include high-achieving agents and struggling agents; The configuration of the learning-optimized agent includes: the density of the prior knowledge base is set to mount the full pre-order course vector library, the decay coefficient of working memory is set to low forgetting, and the maximum reasoning depth is set to support multi-hop reasoning. The configuration of the learning-struggling agent includes: setting the density of the prior knowledge base to be mounted with a sparsed pre-course vector library, setting the working memory decay coefficient to high forgetting, limiting the attention focus window size to sliding window mode, and setting the maximum reasoning depth to single-step reasoning.

4. The method for dynamic adaptive evaluation of digital teaching materials using multidimensional cognitive state vector simulation agents according to claim 1, characterized in that, The S3 includes: The virtual learning agent is aided in understanding by using retrieval-enhanced generation technology, and a knowledge mask matrix is ​​constructed to restrict the virtual learning agent's access to external knowledge. For each group of virtual learning agents, obtain the requests from the virtual learning agents to retrieve nodes in the teaching knowledge graph according to the topological sorting, query the existence status of the nodes in the prior knowledge base, if the query result is not null, output the understanding feedback of the node and the confusion level and thought chain in the process of generating the understanding feedback based on the query result; if the query result is null, force the virtual learning agents to infer the meaning of the nodes only using the context of the current digital teaching materials, if the inference is successful, output the understanding feedback of the node and the confusion level and thought chain in the process of generating the understanding feedback, if the inference fails, trigger an alarm.

5. The method for evaluating the dynamic adaptability of multidimensional cognitive state vector simulation agent digital teaching materials according to claim 4, characterized in that, The S3 further includes: Determine whether the image node plays a cognitive scaffolding role in the process of generating comprehension feedback, and obtain the judgment result that the image node is a valid or invalid cognitive scaffold. Among these, determining whether image nodes act as cognitive scaffolds in the process of generating comprehension feedback includes: The virtual learning agent generates the level of confusion and thought process based on the text nodes during the comprehension feedback process; The virtual learning agent generates new levels of confusion and thought processes based on text and image nodes during the comprehension feedback process; By comparing the level of confusion and the thought chain with the new level of confusion and thought chain, it can be determined whether the image nodes play a cognitive scaffolding role in the process of generating comprehension feedback.

6. The method for dynamic adaptive evaluation of multidimensional cognitive state vector simulation agent digital teaching materials according to claim 1, characterized in that, The cognitive impairment index in S4 is shown in the following formula (1): （1） In the formula, Indicates cognitive impairment index, This represents the total number of nodes in the graph. , , Indicates the weighting coefficient. Indicates an indicator function, Indicates the first The perplexity of each node, Indicates the preset threshold. This indicates that the virtual learning agent is at the node. Number of self-correction attempts at the location This indicates the maximum number of self-corrections allowed by default. The cosine similarity represents the feature vectors of the node's graph and text. Represents the text feature vector of a node. This represents the feature vector of the node image.

7. The method for dynamic adaptive evaluation of multidimensional cognitive state vector simulation agent digital teaching materials according to claim 1, characterized in that, The method further includes: Nodes whose confusion level exceeds a preset threshold or whose thought chain cannot be closed within the maximum reasoning depth step are marked as cognitive breakpoints. The difference between the thought chains of high-achieving agents and low-achieving agents is calculated, and repair suggestions for cognitive breakpoints are generated based on the difference. The repair suggestions include supplementary definitions, analogy examples, and logical connectors.

8. A multidimensional cognitive state vector simulation agent digital textbook dynamic adaptive evaluation device, wherein the multidimensional cognitive state vector simulation agent digital textbook dynamic adaptive evaluation device is used to implement the multidimensional cognitive state vector simulation agent digital textbook dynamic adaptive evaluation method as described in any one of claims 1-7, characterized in that, The device includes: The graph construction module is used to obtain the source files of the digital teaching materials to be evaluated, extract the teaching entities and semantic relationships between entities in the source files of the digital teaching materials through a multimodal parsing model, and construct a directed weighted teaching knowledge graph. The agent factory module is used to initialize multiple sets of virtual learning agents based on a pre-trained large language model, and to configure a differentiated cognitive state vector for each set of virtual learning agents. The cognitive state vector includes the density of the prior knowledge base, the decay coefficient of working memory, the maximum inference depth, and the size of the attention focusing window. The simulation and deduction module is used to drive each group of virtual learning agents to traverse the teaching knowledge graph according to the topological sorting by using the knowledge boundary masking mechanism, and to obtain the understanding feedback of each group of virtual learning agents for each node, as well as the confusion level and thought chain in the process of generating understanding feedback. The evaluation calculation module is used to calculate the cognitive hindrance index and adaptive distribution curve of the digital textbook source file based on the confusion level and thought chain of all virtual learning agents for all nodes, using a weighted statistical algorithm, and generate an evaluation report.

9. A dynamic adaptability evaluation device for digital teaching materials, characterized in that, The digital textbook dynamic adaptability evaluation device includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 7.

Images