Student ability diagnosis method and related device, storage medium

Abstract

The present application provides a student ability diagnosis method and related equipment and storage medium, relates to the technical field of education evaluation, and the student ability diagnosis method comprises the following steps: step one, multi-modal data acquisition; step two, data modeling; step three, hierarchical ability factor decoupling calculation; step four, generalization calibration; and step five, diagnosis visualization. The present application constructs a multi-modal cross-attention mapping model based on screen interaction traces, physiological perception traces and speech thinking traces, inputs the aligned multi-modal cognitive trace tensor, outputs the cognitive input degree label corresponding to each problem solving step, diagnoses the explainability, all weak points of ability can be traced back to the corresponding problem solving trace segment, and the black box diagnosis is completely eliminated.

Description

Technical Field

[0001] This invention relates to the field of educational assessment technology, and in particular to methods for diagnosing student abilities and related equipment and storage media. Background Technology

[0002] Most current mainstream student ability diagnostic schemes rely solely on the correctness of the final answer as the only input, outputting only a binary conclusion of knowledge point mastery / lack of mastery. They fail to uncover the implicit cognitive state of students during the problem-solving process, resulting in three major flaws: First, the interpretability of diagnostic results is poor. The weaknesses in ability output by traditional black-box models cannot be traced back to their source, and teachers and students cannot clearly know which step went wrong. Second, cold-start efficiency is low. Effective diagnosis requires accumulating a large amount of standardized test answer data, and reliable conclusions are almost impossible to output when new students have only answered 1-2 questions. Third, the ability dimension is singular. It can only judge the degree of mastery of knowledge points and cannot identify deeper ability deficiencies such as careless calculation, logical breaks, and insufficient metacognitive regulation.

[0003] The existing published patent CN202111650091.X uses the approach of breaking down the answer steps and matching the standard answer to carry out diagnosis. However, it still does not deviate from the technical framework centered on the final answer result. It cannot capture special cognitive scenarios such as students guessing the right answer in the problem-solving process but taking a roundabout route but ultimately getting the correct result. Therefore, the diagnostic accuracy has obvious defects.

[0004] Therefore, it is necessary to provide a new method for diagnosing student abilities, as well as related equipment and storage media, to solve the above-mentioned technical problems. Summary of the Invention

[0005] To address the technical problem of significant deficiencies in diagnostic accuracy, this invention provides a student ability diagnosis method, related equipment, and storage media. It completely departs from the traditional approach centered on correct or incorrect answers. By collecting multimodal non-outcome cognitive traces throughout the entire problem-solving process, it achieves highly interpretable, low cold-start threshold, and multidimensional accurate ability diagnosis.

[0006] The student ability diagnosis method provided by this invention includes the following complete execution steps: Step 1: Full-link acquisition of multimodal cognitive traces Three types of non-outcome cognitive trace data were collected simultaneously throughout the entire process of students completing problem-solving tasks: the first type is screen interaction traces, including mouse trajectory pause points, drag operation sequences, page review duration distribution, and handwritten handwriting coordinate point sequences in the draft area; the second type is physiological perception traces, including eye-tracking hotspot distribution and facial attention score time series values ​​collected by ordinary consumer-grade cameras; the third type is phonological thought traces, including the phonological transcription of students' thought processes while solving problems. The three types of trace data were aligned on the time axis, with a minimum sampling granularity of 100ms, to generate a multimodal cognitive trace tensor with a unified timestamp. The tensor dimension is [total sampling duration / 0.1, 128], where 128 is the total dimension of multimodal features under a single time slice. Step Two: Cognitive Trace - Problem-Solving Logic Bidirectional Mapping Model A multimodal cross-attention mapping model is constructed, taking the aligned multimodal cognitive trace tensor as input and outputting the cognitive input label corresponding to each problem-solving sub-step. The core of the model adopts a bidirectional cross-attention mechanism, with the following formula: in, This is a feature vector of screen interaction traces. This represents the feature vector of physiological perception traces. This is a feature vector representing speech thought processes. The feature vector dimension is set to 64. A custom mask matrix is ​​used to filter out invalid noise data in the first 3 seconds and the last 2 seconds after the problem-solving task begins. This model can automatically identify three implicit cognitive states of students during the problem-solving process: getting stuck, pausing, jumping around in their thinking, and guessing based on intuition. It does not require the final answer to be the sole criterion for judgment. Step 3: Decoupling Calculation of Hierarchical Capability Factors Students' comprehensive abilities are broken down into three independent and quantifiable ability factors: The first factor is basic calculation ability, scored by statistically analyzing the completeness of calculation steps in the handwriting draft area and the number of pauses in calculation; the second factor is logical reasoning ability, scored by statistically analyzing the frequency of eye-tracking focus jumps between the question stem conditions and reasoning steps and the completeness of the verbal logical chain; the third factor is metacognitive regulation ability, scored by statistically analyzing the percentage of students who actively revisit the question stem and adjust their solution path after getting stuck; the calculation formulas for the three ability factors are as follows: Basic computing power factor ,in This represents the total number of pauses during the computation steps. The total number of calculation steps actually written by the students; Logical deduction ability factor ,in This represents the effective number of jumps between the conditions in the question stem and the derivation steps in the eye-tracking hotspot. This represents the total number of redirects. This is the weight for the completeness of the speech logic chain, with a value range of 0-1. Metacognitive regulation ability factor ,in The percentage of times a solution path was proactively adjusted after getting stuck. For students who waste time on incorrect thinking, The total time spent solving the problem; Step 4: Cross-Scenario Capability Generalization Calibration A cross-scenario transfer calibration mechanism for the same question type is introduced. The student's three-layer ability factor score in the current problem-solving task is dynamically weighted and calibrated with the student's ability factor scores in n questions on the same knowledge point but different scenarios over the past n days. The weights decay exponentially with time intervals. The calibration formula is: in The time interval between the i-th historical question and the current question is in days. The final output is a calibrated three-dimensional ability radar chart, which also marks three types of accurate diagnostic conclusions for students: careless points in calculation, logical breakpoints, and metacognitive control blind spots. Step 5: Generation of Interpretability Visualization of Diagnostic Results The system automatically generates diagnostic reports with links to trace cognitive traces. Clicking on any weakness in the report allows users to jump back to a playback segment of the student's entire problem-solving process during the corresponding time period. This makes the diagnostic results completely traceable and verifiable, completely avoiding the problem of traditional black-box models' diagnostic conclusions being unexplainable.

[0007] Furthermore, in the preprocessing stage of the multimodal cognitive trace tensor in step one, wavelet transform is used to denoise the handwriting coordinate sequence. The db4 wavelet is selected as the mother wavelet, and the decomposition level is set to 3 levels to filter out high-frequency noise caused by hand tremors during handwriting and retain the real writing pauses and trajectory features.

[0008] Furthermore, the custom mask matrix in step two The matrix is ​​an upper triangular matrix. Only the feature weights within the effective time period of the problem-solving task are retained, and the attention weights within the invalid time period are forcibly set to - to avoid irrelevant behavioral data from interfering with the diagnostic results.

[0009] Furthermore, the scores of the three ability factors in step three are all normalized to 0-1. The final diagnostic results are not directly labeled with excellent / good / poor level labels, but instead output the percentile ranking of the student's ability factors in the same grade and knowledge point group, accurately locating the student's relative ability position.

[0010] Furthermore, the cross-scenario transfer calibration mechanism in step four automatically calls the average ability factor distribution of students of the same age as the initial calibration benchmark for new students who have never answered questions on the same knowledge point. It only requires cognitive trace data from one question to complete an effective diagnosis, and the cold start efficiency is 70% higher than the traditional diagnostic scheme based on answer results.

[0011] A second aspect of the present invention provides a student ability diagnostic device based on multimodal cognitive traces, comprising: The multimodal acquisition module is used to simultaneously acquire three types of cognitive trace data: screen interaction, physiological perception, and voice thinking. After timeline alignment, it generates a multimodal cognitive trace tensor. The modeling and calculation module is connected to the multimodal acquisition module to construct a multimodal cross-attention mapping model and identify the cognitive engagement labels of students in the problem-solving process; The hierarchical factor decoupling module is connected to the modeling and calculation module, and independently calculates the quantitative scores of the three-layer ability factors of basic operation, logical deduction and metacognitive regulation. The generalized calibration module, connected to the hierarchical factor decoupling module, performs dynamic calibration of historical data through time exponential decay weights and outputs the final three-dimensional capability diagnostic results. The visualization module, connected to the generalization calibration module, automatically generates and displays a diagnostic report based on the classification and diagnostic conclusions.

[0012] Furthermore, the device is compatible with ordinary cameras and microphones on home computers and tablets, eliminating the need for expensive external peripherals such as eye trackers and pressure pens. It can achieve cognitive state recognition accuracy of no less than 85% on consumer-grade hardware.

[0013] The third aspect of this invention provides a method for storing diagnostic bitstreams of multimodal cognitive traces. The method packages three types of data—desensitized multimodal feature tensors, three-layer ability factor scores, and interpretable diagnostic reports—generated by the method into encrypted bitstreams. These are then encrypted using the national cryptographic algorithm SM4 and stored in a local database. Each time a new diagnostic bitstream is generated, it is automatically associated with the corresponding student's historical bitstream generation ability change trend chain, supporting long-term ability growth retrospective analysis across semesters.

[0014] A fourth aspect of the present invention provides an electronic device comprising a memory and a processor coupled to each other, wherein the memory stores program instructions and the processor is used to execute the program instructions to implement the student ability diagnosis method.

[0015] The fifth aspect of the present invention provides a computer-readable storage medium storing computer program instructions and an encrypted bit stream thereon. When the computer program instructions are executed by a processor, they implement the student ability diagnosis method described above and complete the full-process diagnosis calculation of student abilities.

[0016] Compared with related technologies, the student ability diagnosis method, related equipment, and storage medium provided by this invention have the following beneficial effects: 1. This invention constructs a multimodal cross-attention mapping model based on screen interaction traces, physiological perception traces, and voice thought traces. It takes an aligned multimodal cognitive trace tensor as input and outputs a cognitive input label corresponding to each problem-solving step to diagnose interpretability. All weak points in ability can be traced back to the corresponding problem-solving trace fragments, completely eliminating the need for black-box diagnosis.

[0017] 2. This invention breaks down students' comprehensive abilities into three independent and quantifiable ability factors. The first layer is the basic calculation ability factor, which is scored by statistically analyzing the completeness of calculation steps in the handwriting draft area and the number of calculation pauses. The second layer is the logical deduction ability factor, which is scored by statistically analyzing the frequency of eye-tracking hotspots jumping between the question stem conditions and deduction steps and the completeness of the verbal logical chain. The third layer is the metacognitive regulation ability factor, which is scored by statistically analyzing the proportion of students who actively revisit the question stem and adjust their problem-solving path after getting stuck. The cold start threshold is extremely low, and only the cognitive trace of one question is needed to complete an effective diagnosis. New students do not need to accumulate a large amount of historical answer data.

[0018] 3. This invention dynamically weights and calibrates the student's three-layer ability factor score obtained in the current problem-solving task with the student's ability factor scores in n questions on the same knowledge point but different contexts over the past n days. The weights decay exponentially with time intervals, and the calibrated three-dimensional ability radar chart is output. At the same time, the student's classification diagnosis conclusion is marked. The ability dimensions are richer, and it breaks away from the traditional single evaluation framework of knowledge point mastery. It accurately locates the student's ability deficiencies from three deep dimensions: operation, logic, and metacognition, and provides a more accurate basis for subsequent personalized tutoring. Attached Figure Description

[0019] Figure 1 A structural block diagram of the student ability diagnostic device provided by the present invention; Figure 2 A structural block diagram of screen interaction traces provided by the present invention; Figure 3 A structural block diagram of physiological sensory traces provided by the present invention; Figure 4 A structural block diagram of speech thought traces provided by the present invention; Figure 5 This is a flowchart of the student ability diagnosis method provided by the present invention. Detailed Implementation

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0021] Please refer to the following: Figure 1 , Figure 2 , Figure 3 , Figure 4 as well as Figure 5 ,in, Figure 1 A structural block diagram of the student ability diagnostic device provided by the present invention; Figure 2 A structural block diagram of screen interaction traces provided by the present invention; Figure 3 A structural block diagram of physiological sensory traces provided by the present invention; Figure 4 A structural block diagram of speech thought traces provided by the present invention; Figure 5 This is a flowchart of the student ability diagnosis method provided by the present invention.

[0022] Example 1 In the specific implementation process, such as Figures 1-5 As shown, the student ability diagnosis method provided by this invention includes the following complete execution steps: Step 1: Full-link acquisition of multimodal cognitive traces Three types of non-outcome cognitive trace data were collected simultaneously throughout the entire process of students completing problem-solving tasks: the first type is screen interaction traces, including mouse trajectory pause points, drag operation sequences, page review duration distribution, and handwritten handwriting coordinate point sequences in the draft area; the second type is physiological perception traces, including eye-tracking hotspot distribution and facial attention score time series values ​​collected by ordinary consumer-grade cameras; the third type is phonological thought traces, including the phonological transcription of students' thought processes while solving problems. The three types of trace data were aligned on the time axis, with a minimum sampling granularity of 100ms, to generate a multimodal cognitive trace tensor with a unified timestamp. The tensor dimension is [total sampling duration / 0.1, 128], where 128 is the total dimension of multimodal features under a single time slice. Step Two: Cognitive Trace - Problem-Solving Logic Bidirectional Mapping Model A multimodal cross-attention mapping model is constructed, taking the aligned multimodal cognitive trace tensor as input and outputting the cognitive input label corresponding to each problem-solving sub-step. The core of the model adopts a bidirectional cross-attention mechanism, with the following formula: in, This is a feature vector of screen interaction traces. This represents the feature vector of physiological perception traces. This is a feature vector representing speech thought processes. The feature vector dimension is set to 64. A custom mask matrix is ​​used to filter out invalid noise data in the first 3 seconds and the last 2 seconds after the problem-solving task begins. This model can automatically identify three implicit cognitive states of students during the problem-solving process: getting stuck, pausing, jumping around in their thinking, and guessing based on intuition. It does not require the final answer to be the sole criterion for judgment. Step 3: Decoupling Calculation of Hierarchical Capability Factors Students' comprehensive abilities are broken down into three independent and quantifiable ability factors: The first factor is basic calculation ability, scored by statistically analyzing the completeness of calculation steps in the handwriting draft area and the number of pauses in calculation; the second factor is logical reasoning ability, scored by statistically analyzing the frequency of eye-tracking focus jumps between the question stem conditions and reasoning steps and the completeness of the verbal logical chain; the third factor is metacognitive regulation ability, scored by statistically analyzing the percentage of students who actively revisit the question stem and adjust their solution path after getting stuck; the calculation formulas for the three ability factors are as follows: Basic computing power factor ,in This represents the total number of pauses during the computation steps. The total number of calculation steps actually written by the students; Logical deduction ability factor ,in This represents the effective number of jumps between the conditions in the question stem and the derivation steps in the eye-tracking hotspot. This represents the total number of redirects. This is the weight for the completeness of the speech logic chain, with a value range of 0-1. Metacognitive regulation ability factor ,in The percentage of times a solution path was proactively adjusted after getting stuck. For students who waste time on incorrect thinking, The total time spent solving the problem; Step 4: Cross-Scenario Capability Generalization Calibration A cross-scenario transfer calibration mechanism for the same question type is introduced. The student's three-layer ability factor score in the current problem-solving task is dynamically weighted and calibrated with the student's ability factor scores in three questions covering the same knowledge point but in different contexts over the past seven days. The weights decay exponentially with time intervals. The calibration formula is as follows: in The time interval between the i-th historical question and the current question is in days. The final output is a calibrated three-dimensional ability radar chart, which also marks three types of accurate diagnostic conclusions for students: careless points in calculation, logical breakpoints, and metacognitive control blind spots. Step 5: Generation of Interpretability Visualization of Diagnostic Results The system automatically generates diagnostic reports with links to trace cognitive traces. Clicking on any weakness in the report allows users to jump back to a playback segment of the student's entire problem-solving process during the corresponding time period. This makes the diagnostic results completely traceable and verifiable, completely avoiding the problem of traditional black-box models' diagnostic conclusions being unexplainable.

[0023] The following is an example of how to calculate the ability factor: Suppose a student solves the problem as follows: The calculation involves 10 steps and 5 pauses. ; If there are 20 eye saccades, of which 12 are valid, and the speech weight is 0.8, then... ; The route was adjusted 3 times, with a total time of 10 minutes, of which 3 minutes were wasted. The final percentile ranking is dynamically calculated based on the grade-level group data.

[0024] Specifically, in the preprocessing stage of the multimodal cognitive trace tensor in step one, wavelet transform is used to denoise the handwriting coordinate sequence. The db4 wavelet is selected as the mother wavelet, and the decomposition level is set to 3 levels to filter out high-frequency noise caused by hand tremors during handwriting and retain the real writing pauses and trajectory features.

[0025] Specifically, the custom mask matrix in step two The matrix is ​​an upper triangular matrix. Only the feature weights within the effective time period of the problem-solving task are retained, and the attention weights within the invalid time period are forcibly set to - to avoid irrelevant behavioral data from interfering with the diagnostic results.

[0026] Specifically, the scores of the three ability factors in step three are all normalized to 0-1. The final diagnostic results are not directly labeled with excellent / good / poor levels, but instead output the percentile ranking of the student's ability factors in the same grade and knowledge point group, accurately locating the student's relative ability position.

[0027] Specifically, the cross-scenario transfer calibration mechanism in step four automatically calls the average ability factor distribution of students of the same age as the initial calibration benchmark for new students who have never answered questions on the same knowledge point. It only requires cognitive trace data from one question to complete an effective diagnosis, and the cold start efficiency is 70% higher than the traditional diagnostic scheme based on answer results.

[0028] Example 2

[0029] A second aspect of the present invention provides a student ability diagnostic device based on multimodal cognitive traces, comprising: The multimodal acquisition module is used to simultaneously acquire three types of cognitive trace data: screen interaction, physiological perception, and voice thinking. After timeline alignment, it generates a multimodal cognitive trace tensor. The modeling and calculation module is connected to the multimodal acquisition module to construct a multimodal cross-attention mapping model and identify the cognitive engagement labels of students in the problem-solving process; The hierarchical factor decoupling module is connected to the modeling and calculation module, and independently calculates the quantitative scores of the three-layer ability factors of basic operation, logical deduction and metacognitive regulation. The generalized calibration module, connected to the hierarchical factor decoupling module, performs dynamic calibration of historical data through time exponential decay weights and outputs the final three-dimensional capability diagnostic results. The visualization module, connected to the generalization calibration module, automatically generates and displays a diagnostic report based on the classification and diagnostic conclusions.

[0030] Specifically, the device is compatible with ordinary cameras and microphones on home computers and tablets, without the need for expensive external peripherals such as eye trackers and pressure pens. It can achieve cognitive state recognition with an accuracy of no less than 85% on consumer-grade hardware.

[0031] According to embodiments of the present invention, a computing device that can be used to implement the above method includes a processor and a memory; The processor can be a multi-core processor or include multiple processors. In some embodiments, the processor may include a general-purpose main processor and one or more special coprocessors, such as a graphics processing unit (GPU), a digital signal processor (DSP), etc. In some embodiments, the processor may be implemented using custom circuitry, such as an application-specific integrated circuit (ASIC) or a field-programmable gate array (FPGA).

[0032] Memory can include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage devices. ROM can store static data or instructions required by the processor or other modules of the computer. Permanent storage devices can be read-write storage devices. Permanent storage devices can be non-volatile storage devices that retain stored instructions and data even when the computer is powered off. In some embodiments, permanent storage devices use mass storage devices (e.g., magnetic or optical disks, flash memory) as permanent storage devices. In other embodiments, permanent storage devices can be removable storage devices (e.g., floppy disks, optical drives). System memory can be read-write storage devices or volatile read-write storage devices, such as dynamic random access memory. System memory can store some or all of the instructions and data required by the processor during operation. Furthermore, memory can include any combination of computer-readable storage media, including various types of semiconductor memory chips (DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and disks and / or optical disks can also be used. In some implementations, the memory may include removable storage devices that are readable and / or writable, such as laser discs (CDs), read-only digital versatile optical discs (e.g., DVD-ROMs, dual-layer DVD-ROMs), read-only Blu-ray discs, ultra-high density optical discs, flash memory cards (e.g., SD cards, mini SD cards, Micro-SD cards, etc.), magnetic floppy disks, etc. Computer-readable storage media do not include carrier waves and transient electronic signals transmitted wirelessly or via wired connections.

[0033] It should be understood that, unless otherwise expressly stated herein, there is no strict order restriction on the execution of the above steps, and these steps may be executed in other orders. Moreover, at least some steps in the processes involved in the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but may be executed at different times. The execution order of these steps or stages is not necessarily sequential, but may be performed alternately or in turn with other steps or at least some of the steps or stages in other steps.

[0034] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or basic characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects. The scope of the invention is defined by the appended claims rather than the foregoing description. Therefore, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.

[0035] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment includes only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A student ability diagnosis method, characterized in that, The method includes the following steps: Step 1: Multimodal data acquisition: Simultaneously collect screen interaction traces, physiological perception traces, and voice thinking traces throughout the entire process of students completing the problem-solving task, and preprocess the collected data to generate a multimodal cognitive trace tensor with a unified timestamp. Step 2, Data Modeling: Construct a multimodal cross-attention mapping model based on screen interaction traces, physiological perception traces, and voice thought traces. Input the aligned multimodal cognitive trace tensor and output the cognitive input label corresponding to each problem-solving step. Step 3: Decoupled Calculation of Hierarchical Ability Factors: Students' comprehensive abilities are broken down into three independent and quantifiable ability factors. The first layer is the basic calculation ability factor, which is scored by statistically analyzing the completeness of calculation steps in the handwriting draft area and the number of calculation pauses. The second layer is the logical reasoning ability factor, which is scored by statistically analyzing the frequency of eye-tracking hotspots jumping between the conditions in the question stem and the reasoning steps, and the completeness of the logical chain in the spoken language. The third layer is the metacognitive regulation ability factor, which is scored by statistically analyzing the percentage of students who actively revisit the question stem and adjust their problem-solving path after getting stuck. Step 4, Generalization Calibration: The student's three-layer ability factor score in the current problem-solving task is dynamically weighted and calibrated with the student's ability factor scores in n questions on the same knowledge point but different contexts in the past n days. The weights decay exponentially with time intervals. The calibrated three-dimensional ability radar chart is output, and the student's classification diagnosis conclusion is marked. Step 5: Diagnostic Visualization: Automatically generate and display a diagnostic report based on the classification and diagnostic conclusions.

2. The student ability diagnosis method according to claim 1, characterized in that, The screen interaction traces include mouse trajectory pause points, drag operation sequences, page playback duration distribution, and handwritten handwriting coordinate point sequences in the draft area. The physiological perception traces include the distribution of eye-tracking hotspots and the time-series values ​​of facial attention scores captured by the camera. The aforementioned voice thought traces include the voice-to-text stream of students' verbal explanations of their thought processes while solving problems.

3. The student ability diagnosis method according to claim 1, characterized in that, In the preprocessing stage of the multimodal cognitive trace tensor in step one, wavelet transform is used to denoise the handwriting coordinate sequence. The db4 wavelet is selected as the mother wavelet, and the decomposition level is set to 3 levels to filter out high-frequency noise caused by hand tremors during handwriting and retain the real writing pauses and trajectory features.

4. The student ability diagnosis method according to claim 1, characterized in that, In step three, the scores of the three ability factors are all normalized to 0-1. The final diagnostic result outputs the student's percentile ranking of ability factors in the same grade and knowledge point group, accurately locating the student's relative ability position.

5. A student ability diagnostic device based on multimodal cognitive traces, applicable to the student ability diagnostic method according to any one of claims 1-4, characterized in that, include: The multimodal acquisition module is used to simultaneously acquire three types of cognitive trace data: screen interaction, physiological perception, and voice thinking. After timeline alignment, it generates a multimodal cognitive trace tensor. The modeling and calculation module is connected to the multimodal acquisition module to construct a multimodal cross-attention mapping model and identify the cognitive engagement labels of students in the problem-solving process; The hierarchical factor decoupling module is connected to the modeling and calculation module, and independently calculates the quantitative scores of the three-layer ability factors of basic operation, logical deduction and metacognitive regulation. The generalized calibration module, connected to the hierarchical factor decoupling module, performs dynamic calibration of historical data through time exponential decay weights and outputs the final three-dimensional capability diagnostic results. The visualization module, connected to the generalization calibration module, automatically generates and displays a diagnostic report based on the classification and diagnostic conclusions.

6. A method for storing a multimodal cognitive trace diagnostic bitstream, characterized in that, The three types of data generated by the method described in any one of claims 1-4—the anonymized multimodal feature tensor, the three-layer ability factor score, and the interpretability diagnostic report—are packaged into an encrypted bitstream. After being encrypted using the national cryptographic SM4 algorithm, the bitstream is stored in a local database. Each time a new diagnostic bitstream is generated, it is automatically associated with the corresponding student's historical bitstream generation ability change trend chain, supporting long-term ability growth retrospective analysis across semesters.

7. An electronic device, characterized in that, The method includes a memory and a processor coupled to each other, wherein the memory stores program instructions and the processor executes the program instructions to implement the student ability diagnosis method according to any one of claims 1-4.

8. A computer-readable storage medium storing thereon computer program instructions and the encrypted bit stream of claim 6, characterized in that, When the computer program instructions are executed by the processor described in claim 7, they implement the student ability diagnosis method described in any one of claims 1-4, and complete the full-process diagnosis calculation of student abilities.

Images