Experiment report generation method and device based on large model

Abstract

The invention discloses an experiment report generation method and device based on a large model, and relates to the technical field of natural language processing, and the method comprises the steps: extracting experiment features from an experiment record, and constructing a three-dimensional ethical constraint matrix; implanting ethical supervision and scientific derivation double-unit channels in the large model, and jointly training the large model through compliance samples and violation samples; designing a sample primary verification engine, performing sample primary screening, scoring and confirmation, and outputting an ethical compliance and scientific rationality report; backstepping ethical supervision and scientifically deducing a dependency relationship of double units through the attention weight, and constructing a three-dimensional map; and positioning and correcting the high-risk sample nodes by using the three-dimensional atlas, generating a correction log, integrating the correction log to the three-dimensional ethical constraint matrix, and outputting a final correction report. According to the method, double-unit channels of ethical supervision and scientific derivation are implanted into the BERT model, so that double precise supervision on ethical compliance and scientific rationality in an experimental report generation process is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and particularly to a method and device for generating experimental reports based on large models. Background Art

[0002] In recent years, with the rapid development of artificial intelligence, especially the breakthroughs in the fields of deep learning and natural language processing (NLP), text generation methods based on large models have shown significant application potential in multiple fields. In the field of experimental report generation, traditional methods mostly rely on template-based automated tools, which can achieve the output of structured content to a certain extent. Pretrained language models (such as BERT, GPT) have been introduced into the experimental report generation task, improving the automation level and content quality of the report through semantic understanding and generation capabilities of experimental records. In addition, research combining ethical constraints and risk assessment has gradually emerged, aiming to ensure the compliance and scientificity of the generated content. However, these technologies mostly focus on single-dimensional text generation or compliance detection, lacking the modeling and comprehensive verification of multi-dimensional features of the experimental process.

[0003] Although the existing technologies have made certain progress in automated generation, there are still several deficiencies. First, the traditional methods have limited ability to extract ethically sensitive attributes and potential risks in experimental records, making it difficult to comprehensively evaluate the compliance and safety of experiments and easily resulting in key ethical issues being omitted in the generated reports. Second, although the existing large models have powerful generation capabilities, they lack a dual supervision mechanism for ethics and scientific logic, resulting in possible compliance deviations or logical errors in the output content. In addition, the existing technologies lack sufficient support for the interpretability of the model decision-making process, making it difficult for users to trace the basis of the generated content and make effective corrections. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method for generating experimental reports based on large models to solve the problems of ethical compliance and scientific rationality in experimental report generation.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for generating an experimental report based on a large model, which includes extracting experimental features from experimental records, constructing a three-dimensional ethical constraint matrix, and outputting an experimental risk rating and rule mapping; implanting a dual-unit channel of ethical supervision and scientific derivation in the large model, and jointly training the large model with compliant samples and non-compliant samples to obtain a bimodal feature set; designing a sample primary verification engine to perform preliminary screening, scoring, and confirmation of samples, and outputting a report on ethical compliance and scientific rationality; inferring the dependence relationship between the dual units of ethical supervision and scientific derivation through attention weights, constructing a three-dimensional map, and dynamically linking the three-dimensional ethical constraint matrix; using the three-dimensional map to locate and correct high-risk sample nodes, generating a correction log and integrating it into the three-dimensional ethical constraint matrix, and outputting a final correction report.

[0008] As a preferred embodiment of the method for generating an experimental report based on a large model according to the present invention, wherein: the steps of extracting experimental features from experimental records, constructing a three-dimensional ethical constraint matrix, and outputting an experimental risk rating and rule mapping are specifically as follows.

[0009] The experimental records include operation logs, instrument parameters, and sample identifications.

[0010] Use natural language processing tools to extract experimental features.

[0011] Construct a three-dimensional ethical constraint framework, parse experimental features in combination with predefined rules, and locate the experimental features in the three-dimensional ethical constraint framework.

[0012] Obtain multiple ethical guidelines from an external rule set, number each ethical guideline to form multiple rules, and map them to the three-dimensional ethical constraint framework.

[0013] Integrate experimental features and rule mappings through a priority rule method, output an experimental risk rating, and generate a three-dimensional ethical constraint matrix.

[0014] As a preferred embodiment of the method for generating an experimental report based on a large model according to the present invention, wherein: the steps of implanting a dual-unit channel of ethical supervision and scientific derivation in the large model, and jointly training the large model with compliant samples and non-compliant samples to obtain a bimodal feature set are specifically as follows.

[0015] Select a pre-trained BERT model as the basic framework of the large model.

[0016] Implant an ethical supervision unit and a scientific derivation unit into the output layer of the BERT model.

[0017] Extract experimental text data from operation logs and sample identifications through NLP tools, and extract experimental operation data from operation logs and instrument parameters.

[0018] The three-dimensional ethical constraint matrix is embedded into the fully connected layer of the ethical supervision unit through the attention mechanism for training, and the actual compliance score is output as the basis for evaluating the violation probability of the sample;

[0019] The scientific knowledge base is obtained through external scientific resources and loaded into the scientific derivation unit, and the logical consistency score is output as the basis for verifying the logical error rate of the sample;

[0020] Initial attention weights are assigned to the ethical supervision unit and the scientific derivation unit;

[0021] Compliant samples and non-compliant samples are obtained from historical experiment reports;

[0022] The compliant samples and non-compliant samples are divided into a training set and a validation set according to a ratio;

[0023] The training set is input into the BERT model for training, and the attention weights are adjusted until the ethical supervision unit and the scientific derivation unit reach a balance;

[0024] The sample violation probability and the attention weight distribution are output and combined into a bimodal feature set.

[0025] As a preferred solution of the experimental report generation method based on a large model according to the present invention, wherein: the design sample primary verification engine is configured to perform preliminary screening, scoring, and confirmation of the sample, and output an ethical compliance and scientific rationality report. The specific steps are as follows:

[0026] The sample violation probability is extracted from the bimodal feature set, a violation threshold is set, and high-risk samples exceeding the violation threshold are preliminarily screened out as the first-level review;

[0027] The BERT model is used to perform text analysis on the high-risk samples of the first-level review. The experimental text data and experimental operation data are input, and the compliance score and logical consistency score of each high-risk sample are generated as the second-level review;

[0028] A comprehensive review is performed on the high-risk samples of the second-level review. The three-dimensional ethical constraint matrix is used to compare the sample violation probability, compliance score, and logical consistency score to determine the non-compliant samples and generate a non-compliance conclusion as the third-level review;

[0029] Based on the comprehensive first-level review, second-level review, and third-level review, an ethical compliance and scientific rationality report is output.

[0030] As a preferred solution of the experimental report generation method based on a large model according to the present invention, wherein: the dependency relationship between the two units of ethical supervision and scientific derivation is inferred through the attention weights, a three-dimensional map is constructed, and the three-dimensional ethical constraint matrix is dynamically linked. The specific steps are as follows:

[0031] Infer the dependency relationship between the ethical supervision unit and the scientific derivation unit in the BERT model using the attention weight distribution, compliance score, and logical consistency score;

[0032] Traverse the bimodal feature set and high-risk samples, integrate each high-risk sample, calculate the contribution weights of the ethical supervision unit and the scientific derivation unit, and generate a dependency table;

[0033] Sort the experimental text data and experimental operation data by time to generate the order of experimental steps;

[0034] Load the BERT model, input the dependency table and the three-dimensional ethical constraint matrix, project the attention weight distribution, compliance score, logical consistency score, and the order of experimental steps into a three-dimensional coordinate system, and dynamically link the numbering rules in the three-dimensional ethical constraint matrix to output a three-dimensional map.

[0035] As a preferred solution of the experimental report generation method based on a large model according to the present invention, wherein: the three-dimensional coordinate system means that the X-axis is the order of experimental steps, the Y-axis is the compliance score and the logical consistency score, and the Z-axis is the attention weight and the sample violation probability.

[0036] As a preferred solution of the experimental report generation method based on a large model according to the present invention, wherein: the method of using the three-dimensional map to locate and correct high-risk sample nodes, generate a correction log and integrate it into the three-dimensional ethical constraint matrix, and output a final correction report, the specific steps are as follows

[0037] Analyze the contribution of the attention weight distribution to high-risk samples through the bimodal feature set of the BERT model and the three-dimensional ethical constraint matrix, and mark the high-risk sample nodes;

[0038] Locate the high-risk sample nodes in the three-dimensional map, analyze the dependency relationship of the high-risk sample nodes, propose correction suggestions, and output a correction log;

[0039] Integrate the correction log into the three-dimensional ethical constraint matrix, update the rule mapping of the three-dimensional ethical constraint matrix, and output a final correction report.

[0040] In a second aspect, the present invention provides an experimental report generation device based on a large model, including a risk module, configured to extract experimental features from experimental records, construct a three-dimensional ethical constraint matrix, and output an experimental risk rating and a rule mapping; a training module, configured to implant a dual-unit channel of ethical supervision and scientific derivation in the large model, and jointly train the large model with compliant samples and non-compliant samples to obtain a bimodal feature set; a design module, configured to design a sample primary verification engine, perform preliminary screening, scoring, and confirmation of samples, and output an ethical compliance and scientific rationality report; a construction module, configured to inversely deduce the dependency relationship between the dual units of ethical supervision and scientific derivation through attention weights, construct a three-dimensional map, and dynamically link the three-dimensional ethical constraint matrix; a correction module, configured to use the three-dimensional map to locate and correct high-risk sample nodes, generate a correction log and integrate it into the three-dimensional ethical constraint matrix, and output a final correction report.

[0041] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the experimental report generation method based on a large model as described in the first aspect of the present invention is implemented.

[0042] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the experimental report generation method based on a large model as described in the first aspect of the present invention is implemented.

[0043] The beneficial effects of the present invention are as follows: by implanting a dual-unit channel of ethical supervision and scientific derivation in the BERT model, the present invention realizes the dual-precision supervision of ethical compliance and scientific rationality in the process of experimental report generation. The ethical supervision unit generates a compliance score by embedding a three-dimensional ethical constraint matrix and combining experimental text data, effectively evaluating whether the sample meets regulatory and ethical requirements, ensuring that gene editing experiments have been approved by the state, and avoiding generating non-compliant content. The scientific derivation unit loads a scientific knowledge base, analyzes experimental operation data, and outputs a logical consistency score to verify whether the operation conforms to scientific principles, such as whether the protective measures for high-temperature experiments are reasonable. This dual-unit architecture provides independent ethical and scientific supervision channels for the report content during the generation process, solving the problem that it is difficult for traditional single models to balance compliance and logic. Description of the Drawings

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0045] Figure 1 Flow chart of the experimental report generation method based on large models in Example 1.

[0046] Figure 2 Flow chart of sample review in Example 1. Detailed implementation manners

[0047] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the detailed implementation manners of the present invention in conjunction with the accompanying drawings of the specification.

[0048] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0049] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selectively mutually exclusive embodiment with other embodiments.

[0050] Example 1, referring to Figure 1 and Figure 2 , is the first embodiment of the present invention. This embodiment provides an experimental report generation method based on large models, including the following steps:

[0051] S1. Extract experimental features from the experimental records, construct a three-dimensional ethical constraint matrix, and output an experimental risk rating and a rule mapping.

[0052] Furthermore, the experimental records include operation logs, instrument parameters, and sample identifiers;

[0053] Specifically, the operation logs recorded by the operator during the experiment contain information such as timestamps, operation descriptions, and operator identities;

[0054] The instrument parameters are the operating data of the equipment used in the experiment, such as specific values of temperature, pressure, radiation dose, etc.;

[0055] The sample identifier is the unique identifier of the experimental object and its description;

[0056] Use natural language processing tools to analyze the experimental records line by line to extract experimental features;

[0057] The experimental features include ethical sensitive attributes, experimental risks, and data source reliability markers;

[0058] Specifically, use natural language processing (NLP) tools, such as a BERT-based text analysis model, to parse the text content of the experimental records line by line;

[0059] For the ethically sensitive attributes, identify the descriptions related to ethics through keyword matching and semantic analysis;

[0060] For the experimental risks, identify potential risks based on the operation descriptions and instrument parameters. For example, if the temperature reaches 500°C, it is marked as a high-temperature risk;

[0061] The reliability marking of the data source is to check the qualification information of the record source;

[0062] The ethically sensitive attributes include human tissues, endangered animals, and gene editing. The experimental risks include high-temperature risk, high-pressure risk, and radiation risk;

[0063] The reliability marking of the data source refers to the qualification information of the experimental record source;

[0064] Construct a three-dimensional ethical constraint framework. Through natural language processing tools combined with predefined rules, parse the experimental features and map the experimental features to experimental risk types, experimental stages, and the severity of experimental impacts, and locate them in the three-dimensional ethical constraint framework;

[0065] Specifically, the X-axis / horizontal axis (experimental risk type): Classify risks according to ethically sensitive attributes; Y-axis / vertical axis (experimental stage): Locate according to the log timestamp; Z-axis / vertical axis (severity of impact): Score the severity of impact based on risk discrimination;

[0066] Among them, the predefined rules refer to a set of logics or conditions that have been formulated before the start of the parsing process and are used for mapping experimental features to the framework dimensions;

[0067] Use data processing software (such as Pandas in Python) to generate a three-dimensional ethical constraint framework;

[0068] Define the horizontal axis of the three-dimensional ethical constraint framework as the experimental risk type;

[0069] The experimental risk types include biosafety, data privacy, and academic integrity;

[0070] Map the ethically sensitive attributes and experimental risks to the corresponding types, clarifying the classification basis of the X-axis / horizontal axis;

[0071] Define the vertical axis of the three-dimensional ethical constraint framework as the experimental stage;

[0072] The experimental stage data includes collection, experimental execution, and result analysis;

[0073] Define the vertical axis of the three-dimensional ethical constraint framework as the severity of experimental impact;

[0074] The severity of the experimental impact is divided into five levels;

[0075] Specifically, a 5-level scoring system is adopted. Level 1 is a minor impact, Level 2 is a limited impact, Level 3 is a moderate impact, Level 4 is a high impact, and Level 5 is a severe consequence. The severity of the experimental impact is evaluated according to the experimental risk, and the quantification criteria for the Z-axis / vertical axis are clarified to measure the degree of experimental risk;

[0076] Obtain more than 300 ethical guidelines from external rule sets (such as OSHA safety standards, etc.) and directly convert them into a structured rule library. Each ethical guideline is numbered and mapped to a three-dimensional ethical constraint framework as multiple rules;

[0077] Specifically, collect multiple ethical guidelines, assign a unique number to each guideline, and convert them into a structured format;

[0078] Map multiple ethical guidelines to a three-dimensional ethical constraint framework, such as "rule number → X, experiment execution → Y, Level 4 → Z", and associate multiple ethical guidelines with the three-dimensional ethical constraint framework to form a three-dimensional ethical constraint matrix for reference;

[0079] The structured file is JSON or CSV;

[0080] Integrate the experimental risk types, experimental stages, severity of experimental impact, and rule mappings through the priority rule method, output the experimental risk rating, and generate a three-dimensional ethical constraint matrix with the experimental risk rating and rule mappings;

[0081] First, take the severity of the experimental impact as the initial value. If the rule mapping specifies a specific rating, then follow the rule;

[0082] Experimental risk rating = max(severity of experimental impact, experimental risk rating specified by the rule mapping);

[0083] If the rule clearly defines the experimental risk rating, then use the rule value; otherwise, use the severity of the experimental impact;

[0084] Based on the severity of the experimental impact, if the rule mapping specifies a rating, it is preferred. Output the experimental risk rating and generate a three-dimensional ethical constraint matrix with the experimental risk rating and rule mappings;

[0085] It should be noted that by extracting experimental features from experimental records, constructing a three-dimensional ethical constraint matrix, and outputting experimental risk ratings and rule mappings, the assessment of the ethics and risks of the experimental process is achieved. The NLP tool is used to parse operation logs, instrument parameters, and sample identifiers, and combined with predefined rules, the features are mapped to risk types, experimental stages, and severities to form a structured matrix, providing a compliance basis for subsequent generation. Its role is to comprehensively identify ethical issues and experimental risks in the experiment, and integrate multi-dimensional information through the priority rule method to ensure accurate ratings. Finally, a matrix with experimental risk ratings and rule mappings is generated, enhancing the coverage ability of the report for ethical and safety hazards.

[0086] S2. Implant a dual-unit channel of ethical supervision and scientific derivation in the large model, and jointly train the large model with compliant samples and non-compliant samples to obtain a bimodal feature set.

[0087] Furthermore, select the pre-trained BERT model as the basic framework of the large model;

[0088] Specifically, select the pre-trained BERT-base-uncased model from the HuggingFace model library as the basic framework of the large model;

[0089] Input the maximum sequence length of 512 tokens and the word embedding dimension of 768;

[0090] The output CLS (classification) captures the semantic information of the entire sequence. CLS is a special token in the BERT model used to represent the semantics of the entire input sequence. Its last layer outputs the CLS vector, which is the pooled representation of the sequence and is commonly used in downstream classification tasks;

[0091] Based on the output layer of the BERT model, expand the dual-unit interface, implant the ethical supervision unit and the scientific derivation unit, and define the inputs and outputs of the ethical supervision unit and the scientific derivation unit;

[0092] Specifically, add a fully connected layer and a Sigmoid activation function to the ethical supervision unit respectively to generate a compliance score;

[0093] Add a fully connected layer and a Sigmoid activation function to the scientific derivation unit to generate a logical consistency score;

[0094] Fuse the outputs of the dual units through the attention mechanism, and define the weight parameters w1 (ethics) and w2 (science)

[0095] The CLS vector is input into the dual-unit fully connected layer, and the outputs of the ethical supervision unit and the scientific derivation unit are respectively a single scalar (0-1);

[0096] The input of the ethical supervision unit is experimental text data and a three-dimensional ethical constraint matrix, and the output is a compliance score;

[0097] The input of the scientific derivation unit is experimental operation data, and the output is a logical consistency score;

[0098] Extract experimental text data from operation logs and sample identifiers through NLP tools, and extract experimental operation data from operation logs and instrument parameters;

[0099] The input of the ethical supervision unit is experimental text data and a three-dimensional ethical constraint matrix, and the output is a compliance score;

[0100] Specifically, the ethical supervision unit inputs experimental text data and a three-dimensional ethical constraint matrix;

[0101] Generate a CLS vector through the BERT model, and the fully connected layer combines the rules of the three-dimensional ethical constraint matrix to output a compliance score;

[0102] The embedding layer extracts the CLS vector, and the dimension is 768;

[0103] The fully connected layer activates from ReLU to Sigmoid;

[0104] The output is a compliance score;

[0105] It should be noted that this step is a static definition. It does not explain how the compliance score is generated. It only stipulates the input-output relationship and does not actually perform training or optimization, indicating that the ethical supervision unit will predict compliance based on these inputs;

[0106] The input of the scientific derivation unit is experimental operation data, and the output is a logical consistency score;

[0107] Specifically, the scientific derivation unit inputs experimental operation data;

[0108] Generate a CLS vector through the BERT model, and the fully connected layer combines the experimental operation data to verify logical consistency and outputs a logical consistency score;

[0109] The embedding layer extracts the dimension through the CLS vector;

[0110] The fully connected layer activates from ReLU to Sigmoid;

[0111] The output is a logical consistency score;

[0112] Embed the three-dimensional ethical constraint matrix into the fully connected layer of the ethical supervision unit through the attention mechanism for training. Embed the three-dimensional ethical constraint matrix into the fully connected layer of the ethical supervision unit and output the actual compliance score as the basis for evaluating the violation probability of the sample;

[0113] It should be noted that this step is implemented dynamically, which illustrates that the actual compliance score is generated by embedding the three-dimensional ethical constraint matrix and training;

[0114] Specifically, a compliance score is defined as 0-1, where the closer to 1, the more compliant, and closer to 0, the more non-compliant;

[0115] Logical consistency score, ranging from 0 to 1, the closer to 1, the more logically consistent, and closer to 0, the more logically incorrect;

[0116] For example, if the compliance score is 0.9 and the probability of violation is 0.1 (low violation), then the compliance score of 0.9 is close to 1 and the probability of violation is close to 0, which is a low violation, indicating that the sample is compliant.

[0117] Compliance score 0.2 → Violation probability 0.8 (high violation) → Then the compliance score 0.2 is close to 0 and the violation probability is close to 1, which is a high violation, indicating that the sample violates the rules;

[0118] Logical consistency score 0.8 → sample logical error rate 0.2 (low error) → logical consistency close to 1 (correct), indicating that the sample logic is correct;

[0119] Logical consistency score 0.4 → sample logical error rate 0.6 (high error) → logical consistency close to 0 (error), indicating sample logical errors;

[0120] The three-dimensional ethical constraint matrix is flattened and encoded into a feature vector, which is then fused with the CLS vector through an attention mechanism and input into a fully connected layer. Compliance / violation samples are used to optimize the fully connected layer and attention weights to generate a compliance score.

[0121] This step generates the actual rating capability through the embedding matrix and finally outputs the result;

[0122] Obtain a scientific knowledge base through external scientific resources (physical law database), convert the scientific knowledge base into a structured format through NLP tools, load it into the scientific deduction unit, and output a logical consistency score as the basis for verifying the sample logical error rate (i.e., the updated scientific deduction unit parameters);

[0123] Assign initial attention weights to the ethics supervision unit and scientific deduction unit, such as w1(ethics) = 0.5, w2(science) = 0.5, satisfying w1+w2=1;

[0124] Obtain compliant samples (samples of experimental cases approved by national agencies, such as samples based on approved gene editing experiments) and noncompliant samples (samples of experimental cases that have not obtained national agency permission or violate regulations, such as samples based on unauthorized high-temperature experiments) from historical experimental reports (approximately 5,000 reports). Label compliant samples as positive samples and noncompliant samples as negative samples, and indicate the violation points.

[0125] Compliant samples and non-compliant samples are divided into training set (80%) and validation set (20%) in proportion;

[0126] The training set is input into the BERT model for training. The loss function is used to optimize the ethical supervision unit and the scientific deduction unit, and the attention weight is adjusted to achieve a balance between the ethical supervision unit and the scientific deduction unit.

[0127] Specifically, the ethics supervision unit optimizes the compliance score and generates the probability of sample violation (i.e., 1-compliance score);

[0128] Optimize logical consistency scoring in scientific deduction units;

[0129] Use the cross entropy loss function (sample compliance loss + sample consistency loss) to adjust the attention weight to balance;

[0130] The loss function is L = L c +L o , where L is the loss function, which represents the goal of BERT model optimization. BERT model parameters are usually adjusted by minimizing L. c is the loss of the ethical supervision unit, which measures the gap between the compliance score and the true compliance label (such as 1 or 0). For example, if the sample label is 1 for compliance and the prediction is 0.8, L o The loss of the scientific derivation unit is the loss of the scientific loss, which measures the gap between the logical consistency score and the true consistency label. For example, if the sample label is consistent, it is 1, and the prediction is 0.9, L o Calculate the error and dynamically adjust the total attention weight (e.g., to “0.6 for ethics, 0.4 for science”);

[0131] Output the sample violation probability and attention weight distribution and directly combine them into a bimodal feature set (including sample violation probability, experimental text data, experimental operation data, three-dimensional ethical constraint matrix and scientific knowledge base, that is, the trained BERT model), and save it as a JSON format file;

[0132] It should be noted that the sample violation probability is the result of the ethical judgment of each sample (sample level) and is directly used for screening;

[0133] The attention weight distribution is a trained global weight that is paired with each sample to provide explanatory and bimodal information;

[0134] Use the validation set to test the sample violation probability and sample logical error rate, and adjust the attention weight distribution to optimize the BERT model performance;

[0135] The sample logic error rate is an overall performance metric (at the global level) for the validation set. It is not a sample-level feature and cannot provide specific values for each sample. Thus, it is not suitable as content for the atlas. Its function is to test the performance of the BERT model and optimize the attention weight distribution, rather than directly judging samples.

[0136] Specifically, for the sample violation probability: comparing the scores of the ethical supervision unit with the compliance labels to obtain the accuracy rate (e.g., ≥90%). The calculation of the accuracy rate is: the number of correct predictions / the total number.

[0137] For the sample logic error rate: comparing the scores of the scientific derivation unit with the consistency labels to obtain the error ratio (e.g., ≤5%). The calculation of the error rate is: the number of incorrect predictions / the total number.

[0138] It should be noted that by implanting dual units of ethical supervision and scientific derivation in the BERT model, jointly training with compliant and non-compliant samples, and outputting a bimodal feature set, precise dual supervision of ethics and science in the generation of experimental reports has been achieved. The ethical supervision unit generates compliance scores in combination with a three-dimensional ethical constraint matrix, and the scientific derivation unit outputs logical consistency scores relying on a scientific knowledge base, ensuring that the content of the experimental report complies with regulations and scientific principles. Its role is to provide an intelligent compliance and rationality evaluation tool for the scientific research field, applicable to high-demand scenarios such as gene editing and medical experiments. Finally, through the collaboration of the dual units and the optimization of attention weights, the compliance and logical rigor of the experimental report have been significantly improved, avoiding the biases of a single model and generating a reliable bimodal feature set.

[0139] S3. Design a primary verification engine for samples, conduct preliminary screening, scoring, and confirmation of samples, and output reports on ethical compliance and scientific rationality.

[0140] Furthermore, extract the sample violation probability from the bimodal feature set, set a violation threshold, and preliminarily screen out high-risk samples exceeding the violation threshold for first-level review.

[0141] Specifically, extract the sample violation probability of approximately 5000 samples from the bimodal feature set.

[0142] Use the rules in the three-dimensional ethical constraint matrix as a reference, set a violation threshold (for example, set samples with high experimental risk severity or high risk as 0.7, and samples below 0.3 are initially regarded as low risk) to determine the screening criteria.

[0143] Traverse each sample in the bimodal feature set, compare the sample violation probability of each sample with the violation threshold. If it is higher than the violation threshold (e.g., ≥0.7), then mark it as a high-risk sample to be reviewed.

[0144] It should be noted that the violation threshold is set based on the severity grading (levels 1-5) of the three-dimensional ethical constraint matrix. A sample violation probability ≥ 0.7 is defined as the high-risk threshold, corresponding to samples with a severity of level 4 or above, indicating a high impact that requires immediate review; a violation probability ≤ 0.3 is defined as the low-risk threshold, corresponding to samples with a severity of level 2 or below, initially regarded as low-risk, and samples between 0.3-0.7 are marked as potential risks and require further evaluation;

[0145] Output the list of high-risk samples as the first-level review result;

[0146] Use the BERT model to perform text analysis on the high-risk samples of the first-level review. Input the experimental text data and experimental operation data to generate the compliance score and logical consistency score for each high-risk sample as the second-level review;

[0147] Specifically, read the high-risk samples, load the BERT model, input the experimental text data and experimental operation data, concatenate the experimental text data and experimental operation data into a single sequence, and mark 512 for the maximum sequence length of 512 tokens;

[0148] Input the sequence into the BERT model to generate the CLS vector, that is, the CLS vector;

[0149] Input the CLS vector into the fully connected layers of the ethical supervision unit and the scientific derivation unit respectively;

[0150] The ethical supervision unit outputs the compliance score, indicating the probability that the sample complies with the ethical rules;

[0151] The scientific derivation unit outputs the logical consistency score, indicating the scientific rationality of the sample operation;

[0152] Example: Sample A: Compliance score 0.7, logical consistency score 0.9;

[0153] Sample B: Compliance score 0.6, logical consistency score 0.8;

[0154] Generate the compliance score and logical consistency score for each high-risk sample;

[0155] Conduct a comprehensive review of the high-risk samples of the second-level review. Use the three-dimensional ethical constraint matrix to compare the sample violation probability, compliance score, and logical consistency score to determine the violation samples and generate violation conclusions as the third-level review;

[0156] Specifically, read the compliance score and logical consistency score of the high-risk samples from the high-risk samples with high scores in the second-level review;

[0157] Extract the sample violation probability of the corresponding samples from the bimodal feature set;

[0158] Load the scientific knowledge base and the three-dimensional ethical constraint matrix. For each sample, combine the rules in the three-dimensional ethical constraint matrix and the scientific knowledge base, and conduct a secondary review. For example, the violation probability of the first-level sample, the compliance score of the second level, and the logical consistency score of the second level. If the compliance score is lower than the rule requirements, or the sample violation probability is high and violates the rules of the three-dimensional ethical constraint matrix, it is determined as a violation. If the logical consistency score is low or violates the knowledge base rules, further confirm the violation;

[0159] For example:

[0160] Sample A: The sample violation probability is 0.8, the compliance score is 0.7 (lower than the standard), the logical consistency score is 0.9, and it is not approved and is confirmed as a violation;

[0161] Sample B: The sample violation probability is 0.6, the compliance score is 0.9, and the logical consistency score is 0.8. It is compliant and consistent and is not confirmed as a violation;

[0162] Output the generated list of confirmed violation samples and violation conclusions as the results of the third-level review;

[0163] Integrate the first-level review, second-level review, and third-level review, and output an ethical compliance and scientific rationality report (including a list of high-risk samples, sample score details, violation sample conclusions, and compliance sample statistics);

[0164] It should be noted that by designing a sample primary verification engine, extracting the violation probability from the bimodal feature set and conducting a third-level review, and outputting an ethical compliance and scientific rationality report, the accurate verification of the generated content is achieved. High-risk samples are screened using the violation threshold, the BERT model generates compliance and logic scores, and a comprehensive evaluation is conducted in combination with the review and the three-dimensional ethical constraint matrix to ensure the dual ethical and scientific standards of the samples. Its role is to provide reliable quality control for experimental reports, applicable to high-compliance fields such as medicine and biology. Finally, through multi-level verification, the accuracy and credibility of the report are improved, and the deviation of single automated judgment is avoided.

[0165] S4. Through the reverse inference of the dependence relationship between the ethical supervision and scientific derivation dual units by the attention weights, construct a three-dimensional map and dynamically link the three-dimensional ethical constraint matrix.

[0166] Furthermore, use the attention weight distribution, compliance score, and logical consistency score to reverse infer the dependence relationship between the ethical supervision unit and the scientific derivation unit in the BERT model;

[0167] Specifically, map the attention weight distribution in the bimodal feature set to the tokens input to the BERT model;

[0168] Load the pre-trained BERT model and extract the weight parameters of the ethical supervision unit and the scientific derivation unit of the trained BERT model;

[0169] Use the SHAP method to calculate the contribution values of the attention weight distribution to the compliance score and the logical consistency score, and generate a dependency relationship;

[0170] Traverse the bimodal feature set and high-risk samples, integrate the attention weight distribution, compliance score, and logical consistency score of each high-risk sample, and calculate the contribution weights of the ethical supervision unit and the scientific derivation unit to the sample violation probability, compliance score, and logical consistency score, and generate a dependency table;

[0171] Specifically, use a Python script to traverse each high-risk sample;

[0172] Extract the attention weight distribution, compliance score, logical consistency score, and sample violation probability of each sample;

[0173] Use gradient analysis (e.g., torch.autograd automatic differentiation in PyTorch) to calculate the contribution weights of the ethical supervision unit and the scientific derivation unit to the output, integrate the results, and generate a dependency table;

[0174] Sort the experimental text data and experimental operation data by time to generate the order of experimental steps;

[0175] Specifically, sort the experimental text data and the experimental operation data by time to generate the order of experimental steps;

[0176] Load the BERT model, input the dependency table and the three-dimensional ethical constraint matrix, project the attention weight distribution, compliance score, logical consistency score, and the order of experimental steps into a three-dimensional coordinate system, and dynamically link the numbering rules in the three-dimensional ethical constraint matrix to output a three-dimensional map;

[0177] The three-dimensional coordinate system means that the X-axis is the order of experimental steps;

[0178] The Y-axis is the compliance score and the logical consistency score;

[0179] The Z-axis is the attention weight and the sample violation probability;

[0180] It should be noted that by inversely deducing the dependence relationship between the ethical supervision and scientific derivation units through the attention weights, constructing a three-dimensional map and dynamically linking the three-dimensional ethical constraint matrix, the visualization and traceability of the decision-making process are realized. Using the SHAP method and gradient analysis, the scores and weights in the bimodal feature set are integrated to generate a dependence relationship table, which is projected onto a three-dimensional coordinate system, providing an intuitive analysis tool for users. Its role lies in the basis for ethical and scientific judgments. Finally, through the dynamic link matrix rule, the transparency and interpretability of the generation process are enhanced, facilitating the positioning of problem nodes.

[0181] S5. Locate and correct the high-risk sample nodes using the three-dimensional map, generate a correction log and integrate it into the three-dimensional ethical constraint matrix, and output the final correction report.

[0182] Furthermore, through the bimodal feature set of the BERT model and the three-dimensional ethical constraint matrix, analyze the contribution of the attention weight distribution to high-risk samples, locate the feature dependencies of high-risk samples in the three-dimensional map, and mark the high-risk sample nodes that need to be verified.

[0183] Specifically, extract the sample violation probability (such as sample A: 0.85), the attention weight of the ethical supervision unit (such as 0.7), and the attention weight of the scientific derivation unit (such as 0.3) from the bimodal feature set.

[0184] Extract the three-dimensional coordinates of the high-risk sample nodes from the three-dimensional map.

[0185] Based on the dependence relationship table (the contribution weights of the ethical supervision unit and the scientific derivation unit to the sample violation probability), obtain the ethical contribution ratio and the scientific contribution ratio.

[0186] If the ethical weight is 0.7 and the scientific weight is 0.3, then the ethical contribution ratio is 0.7 / (0.7 + 0.3) = 70%, and the scientific contribution ratio is 30%.

[0187] According to the three-dimensional ethical constraint matrix, map the high-risk sample nodes to the rules in the three-dimensional ethical constraint matrix, such as mapping the experimental step sequence to the experimental stage, the compliance score to the impact severity, and the sample violation probability to the risk rating basis.

[0188] If the compliance score of a certain high-risk sample node is lower than the requirement of the rule mapping, it is marked as an ethical rule conflict.

[0189] If the logical consistency score of a certain high-risk sample node is lower than the scientific knowledge base standard, it is marked as a scientific logic conflict.

[0190] Mark the ethical rule conflict and the scientific logic conflict according to the ethical contribution ratio and the scientific contribution ratio.

[0191] Locate high-risk sample nodes in the three-dimensional atlas, analyze the dependency relationships of the high-risk sample nodes, propose corrective suggestions, and output a correction log;

[0192] Specifically, based on the rule mapping in the three-dimensional ethical constraint matrix, match the reasons for violations and generate corrective suggestions;

[0193] Based on the loaded scientific knowledge base, match the types of violations and generate corrective suggestions;

[0194] Check the instrument parameters and experimental stages in the experimental operation data to confirm the contradictions in the types of violations;

[0195] Generate a correction log for each high-risk sample node (including sample ID, experimental execution stage, type of violation, dependency weight, rule mapping, corrective suggestions, and update status such as "pending" or "resolved");

[0196] Integrate the correction log into the three-dimensional ethical constraint matrix, update the rule mapping of the three-dimensional ethical constraint matrix, and output the final correction report;

[0197] Specifically, if the corrective suggestions involve rules not covered by the three-dimensional ethical constraint matrix, extract and assign new numbers from the rules in the three-dimensional ethical constraint matrix, and map them to the X-axis (experimental risk type), Y-axis (experimental stage), and Z-axis (impact severity) of the three-dimensional matrix;

[0198] Recalculate the experimental rating according to the corrective suggestions, and update the three-dimensional ethical constraint matrix according to the priority rule method (such as downgrading the original rating from level 4 to level 3);

[0199] Output the final correction report (including a list of high-risk samples, correction details, update records of the three-dimensional ethical constraint matrix, and compliance statistics, such as the proportion of compliant samples after correction (such as from 75% to 92%);

[0200] It should be noted that through the bimodal feature set of the BERT model and the three-dimensional ethical constraint matrix, combined with the attention weight distribution, accurately locate the feature dependencies of high-risk sample nodes, analyze the reasons for their violations and generate corrective suggestions, mark ethical rule conflicts or scientific logic conflicts based on rule mapping and the scientific knowledge base, generate a correction log and update the three-dimensional ethical constraint matrix, and finally output a correction report to improve the compliance ratio, optimize the experimental risk rating and rule coverage.

[0201] This embodiment also provides an experimental report generation device based on a large model, including: a risk module, configured to extract experimental features from experimental records, construct a three-dimensional ethical constraint matrix, and output an experimental risk rating and a rule mapping; a training module, configured to implant a dual-unit channel of ethical supervision and scientific derivation in the large model, and jointly train the large model with compliant samples and non-compliant samples to obtain a bimodal feature set; a design module, configured to design a sample primary verification engine, perform preliminary screening, scoring, and confirmation of samples, and output an ethical compliance and scientific rationality report; a construction module, configured to reverse-infer the dependency relationship between the dual units of ethical supervision and scientific derivation through attention weights, construct a three-dimensional map, and dynamically link the three-dimensional ethical constraint matrix; a correction module, configured to use the three-dimensional map to locate and correct high-risk sample nodes, generate a correction log and integrate it into the three-dimensional ethical constraint matrix, and output a final correction report.

[0202] This embodiment also provides a computer device applicable to the situation of the experimental report generation method based on a large model, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the experimental report generation method based on a large model as proposed in the above embodiment.

[0203] This computer device may be a terminal, and this computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of this computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of this computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of this computer device may be a touch layer covered on the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0204] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for generating an experimental report based on a large model as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (abbreviated as SRAM), electrically erasable programmable read-only memory (abbreviated as EEPROM), erasable programmable read-only memory (abbreviated as EPROM), programmable read-only memory (abbreviated as PROM), read-only memory (abbreviated as ROM), magnetic memory, flash memory, a magnetic disk, or an optical disc.

[0205] In summary, by implanting a dual-unit channel of ethical supervision and scientific derivation in the BERT model, jointly training with compliant samples and non-compliant samples, outputting compliance scores and logical consistency scores, and generating a bimodal feature set, the present invention realizes the dual-precision supervision of ethical compliance and scientific rationality in the process of generating an experimental report. The ethical supervision unit generates a compliance score by embedding a three-dimensional ethical constraint matrix and combining experimental text data, effectively evaluating whether the sample meets the regulatory and ethical requirements, ensuring that the gene editing experiment has been approved by the state, and avoiding generating non-compliant content. The scientific derivation unit loads a scientific knowledge base, analyzes experimental operation data, and outputs a logical consistency score to verify whether the operation conforms to scientific principles, such as whether the protective measures for a high-temperature experiment are reasonable. This dual-unit architecture provides independent ethical and scientific supervision channels for the report content during the generation process, solving the problem that it is difficult for traditional single models to balance compliance and logic.

[0206] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An experimental report generation method based on a large model, characterized in that: including extracting experimental features from experimental records, constructing a three-dimensional ethical constraint matrix, and outputting an experimental risk rating and rule mapping implanting dual-unit channels of ethical supervision and scientific derivation into a large model, and jointly training the large model with compliant samples and non-compliant samples to obtain a bimodal feature set designing a sample primary verification engine to perform preliminary screening, scoring, and confirmation of samples, and outputting a report on ethical compliance and scientific rationality inferring the dependency relationship between the dual units of ethical supervision and scientific derivation through attention weights, constructing a three-dimensional map, and dynamically linking the three-dimensional ethical constraint matrix using the three-dimensional map to locate and correct high-risk sample nodes, generating a correction log and integrating it into the three-dimensional ethical constraint matrix, and outputting a final correction report 2. The method for generating an experimental report based on a large model according to claim 1, wherein: The specific steps for extracting experimental features from experimental records, constructing a three-dimensional ethical constraint matrix, and outputting an experimental risk rating and rule mapping are as follows The experimental records include operation logs, instrument parameters, and sample identifiers using natural language processing tools to extract experimental features constructing a three-dimensional ethical constraint framework, parsing experimental features in combination with predefined rules, and locating the experimental features in the three-dimensional ethical constraint framework obtaining multiple ethical guidelines from an external rule set, numbering each ethical guideline to form multiple rules, and mapping them to the three-dimensional ethical constraint framework integrating experimental features and rule mapping through the priority rule method, outputting an experimental risk rating, and generating a three-dimensional ethical constraint matrix 3. The method for generating an experimental report based on a large model according to claim 2, wherein: The specific steps for implanting dual-unit channels of ethical supervision and scientific derivation into a large model, and jointly training the large model with compliant samples and non-compliant samples to obtain a bimodal feature set are as follows selecting a pre-trained BERT model as the basic framework of the large model implanting an ethical supervision unit and a scientific derivation unit into the output layer of the BERT model extracting experimental text data from operation logs and sample identifiers through NLP tools, and extracting experimental operation data from operation logs and instrument parameters embedding the three-dimensional ethical constraint matrix into the fully connected layer of the ethical supervision unit through an attention mechanism for training, and outputting an actual compliance score as the basis for evaluating the violation probability of samples obtaining a scientific knowledge base through external scientific resources, loading it into the scientific derivation unit, and outputting a logical consistency score as the basis for verifying the logical error rate of samples assigning initial attention weights to the ethical supervision unit and the scientific derivation unit obtaining compliant samples and non-compliant samples from historical experimental reports dividing the compliant samples and non-compliant samples into a training set and a validation set according to a ratio inputting the training set into the BERT model for training, and adjusting the attention weights until the ethical supervision unit and the scientific derivation unit reach a balance outputting the violation probability of samples and the distribution of attention weights and combining them into a bimodal feature set 4. The method for generating an experimental report based on a large model according to claim 3, wherein: The specific steps for designing a sample primary verification engine to perform preliminary screening, scoring, and confirmation of samples, and outputting a report on ethical compliance and scientific rationality are as follows extracting the violation probability of samples from the bimodal feature set, setting a violation threshold, and preliminarily screening high-risk samples as the first-level review using the BERT model to perform text analysis on the high-risk samples of the first-level review, generating a compliance score and a logical consistency score for each high-risk sample as the second-level review Conduct a comprehensive review of high-risk samples in the secondary review, identify non-compliant samples and generate non-compliance conclusions as the tertiary review; Integrate the primary review, secondary review and tertiary review, and output a report on ethical compliance and scientific rationality.

5. The method for generating an experimental report based on a large model according to claim 4, wherein: The method of inversely inferring the dependence relationship between the ethical supervision and scientific derivation dual units through attention weights, constructing a three-dimensional map, and dynamically linking the three-dimensional ethical constraint matrix is as follows: Use the attention weight distribution, compliance score and logical consistency score to inversely infer the dependence relationship between the ethical supervision unit and the scientific derivation unit in the BERT model; Traverse the bimodal feature set and high-risk samples, integrate each high-risk sample, calculate the contribution weights of the ethical supervision unit and the scientific derivation unit, and generate a dependence relationship table; Sort the experimental text data and experimental operation data according to time to generate the order of experimental steps; Load the BERT model, input the dependence relationship table and the three-dimensional ethical constraint matrix, project the attention weight distribution, compliance score, logical consistency score and the order of experimental steps into a three-dimensional coordinate system, and dynamically link the numbering rules in the three-dimensional ethical constraint matrix to output a three-dimensional map.

6. The method for generating an experimental report based on a large model according to claim 5, wherein: The three-dimensional coordinate system means that the X-axis is the order of experimental steps, the Y-axis is the compliance score and the logical consistency score, and the Z-axis is the attention weight and the sample non-compliance probability.

7. The method for generating an experimental report based on a large model according to claim 6, wherein: The method of using the three-dimensional map to locate and correct high-risk sample nodes, generate a correction log and integrate it into the three-dimensional ethical constraint matrix, and output a final correction report is as follows: Analyze the contribution of the attention weight distribution to high-risk samples through the bimodal feature set of the BERT model and the three-dimensional ethical constraint matrix, and mark the high-risk sample nodes; Locate the high-risk sample nodes in the three-dimensional map, analyze the dependence relationship of the high-risk sample nodes, put forward correction suggestions, and output a correction log; Integrate the correction log into the three-dimensional ethical constraint matrix, update the rule mapping of the three-dimensional ethical constraint matrix, and output a final correction report.

8. An experimental report generation device based on a large model, based on the experimental report generation method based on a large model according to any one of claims 1 to 7, characterized in that: Including: A risk module for extracting experimental features from experimental records, constructing a three-dimensional ethical constraint matrix, and outputting an experimental risk rating and a rule mapping; A training module for implanting ethical supervision and scientific derivation dual unit channels in a large model, and jointly training the large model with compliant samples and non-compliant samples to obtain a bimodal feature set; A design module for designing a sample primary verification engine, performing preliminary screening, scoring and confirmation of samples, and outputting a report on ethical compliance and scientific rationality; A construction module for inversely inferring the dependence relationship between the ethical supervision and scientific derivation dual units through attention weights, constructing a three-dimensional map, and dynamically linking the three-dimensional ethical constraint matrix; A correction module for using the three-dimensional map to locate and correct high-risk sample nodes, generate a correction log and integrate it into the three-dimensional ethical constraint matrix, and output a final correction report.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method for generating an experimental report based on a large model according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the steps of the method for generating an experimental report based on a large model according to any one of claims 1 to 7.

Images