System for monitoring and intervening large model to generate illusion content in real time

Abstract

The invention provides a system for monitoring and intervening a large model to generate illusion content in real time, and relates to the field of electric digital data processing, the system comprises a large model docking module, an input analysis module, an output monitoring module and an intervention modification module, the large model docking module is used for carrying out data docking with a large model system, and the input analysis module is used for analyzing the illusion content. The input analysis module is used for performing content analysis on input information of a large model, the output monitoring module is used for performing illusion monitoring on output content of the large model, and the intervention modification module is used for modifying and adjusting the illusion content; according to the system, input questions of a large model are analyzed firstly, the illusion type prone to occurring is determined on the basis of the input content type, the analysis range is greatly reduced, then the output content is detected and analyzed in a targeted mode, and the accuracy of illusion detection can be effectively improved.

Description

Technical Field

[0001] The invention relates to the field of electronic digital data processing, and in particular to a system for real-time monitoring and intervening in large model generation of hallucination content. Background Art

[0002] With the rapid development and widespread application of large-scale language models, their performance in natural language processing, information generation, and interactive tasks has been significantly improved. However, since the model generates content mainly based on statistical language probability distribution, it has limitations in grasping factual knowledge, and often produces fictional content or logical contradictions. This phenomenon is called "hallucination". Existing research and technology mainly focus on improving the generation ability and accuracy of the model, such as reducing the occurrence of hallucinations by optimizing pre-training data, introducing knowledge graphs, or building more complex network structures. However, these methods cannot completely avoid hallucination content, especially when the user input context is complex or the knowledge is sparse. On the other hand, some detection systems can mark hallucination content, but they often remain in the passive detection stage and lack effective intervention and correction mechanisms for hallucination content.

[0003] Many hallucination detection systems have been developed. After a lot of searching and reference, it is found that the existing hallucination detection systems are such as the system disclosed in the publication number CN118333149B. The general methods of these systems include: obtaining the output of the first language model, dividing the output into multiple units with a certain order according to the preset units, performing named entity recognition on each unit, and obtaining the category of each unit; using the second language model to predict the probability of generating the corresponding unit at each position of the output; calculating the first hallucination degree value corresponding to the output based on each unit and its category and its corresponding probability; generating multiple questions based on the output and the category of each unit, answering each question using the first language model, and obtaining the answer to each question; calculating the second hallucination degree value corresponding to the output based on the output and the answer to each question; judging whether the output is a hallucination generated by the first language model based on the first hallucination degree value and the second hallucination degree value. However, this system only performs hallucination detection on the output of the large model, which is not targeted, and there is room for improvement in the success rate of hallucination detection. Summary of the invention

[0004] The purpose of the present invention is to propose a system for real-time monitoring and intervening in the generation of hallucination content by a large model in view of the existing deficiencies.

[0005] The present invention adopts the following technical solution: A real-time monitoring and intervention system for large model generated hallucination content, including a large model docking module, an input parsing module, an output monitoring module and an intervention modification module; The large model docking module is used to perform data docking with the large model system, the input parsing module is used to perform content parsing on the input information of the large model, the output monitoring module is used to perform hallucination monitoring on the output content of the large model, and the intervention modification module is used to modify and adjust the hallucination content; The large model docking module includes an input detection unit, an output interception unit and an output feedback unit. The input detection unit is used to detect the input event of the large model and obtain the input content. The output interception unit is used to intercept the output information of the large model. The output feedback unit is used to feed back the actual output information to the large model. The input parsing module includes an information classification unit, an illusion mapping unit and a rule output unit, wherein the information classification unit is used to classify and identify the input information, the illusion mapping unit is used to map the classified information to the illusion type, and the rule output unit is used to send the illusion judgment rule to the output monitoring module; The output monitoring module includes a content management unit, a rule execution unit and a result output unit, wherein the content management unit is used to manage the data content that needs to be hallucinated, the rule execution unit performs judgment analysis on the data content based on the judgment rule, and the result output unit is used to send the judgment result to the intervention modification module; The intervention modification module includes a result diversion unit, an illusion modification unit and a content output unit. The result diversion unit is used to divert the judgment result, the illusion modification unit is used to modify and adjust the illusion content, and the content output unit is used to send the final content information to the output feedback unit.

[0006] Further, the rule output unit includes an illusion rule base, a rule matching processor and a rule packaging processor, wherein the illusion rule base is used to store rule information for judging illusions, the rule matching processor matches corresponding rule information from the illusion rule base based on the illusion type, and the rule packaging processor is used to package all matched rule information and output them; The content management unit includes a detection content register, a detection content update processor and a detection control processor. The detection content register creates a storage area based on the hallucination intervention project to store the output content of the large model. The detection content update processor is used to update the content information in the storage area. The detection control processor organizes the content information in the storage area into an execution data packet for hallucination detection at each update. The execution data packet includes environment data and target data, the environment data is all output content of the hallucination intervention project at the current time point, and the target data is the sentence content newly intercepted by the hallucination intervention project.

[0007] Furthermore, the rule execution unit includes an environment assessment processor, a rule verification processor and a multidimensional scoring processor. The environment assessment processor is used to evaluate and analyze the environment data. The rule verification processor verifies and analyzes the target data based on the rule information and the environment assessment results. The multidimensional scoring processor scores the target data for satisfying the illusion based on the verification results of multiple rules.

[0008] Furthermore, the process of the rule verification processor performing verification analysis on the target data includes the following steps: S1. Obtaining the evaluation results of environmental data and target data; S2, obtaining a rule information; S3, filtering out evaluation items and corresponding evaluation values ​​in the evaluation results based on the rule information; S4. Calculate the check value Vc under the rule information according to the following formula: ; Among them, n is the number of selected evaluation items, Am 1 (i) represents the evaluation value of the i-th evaluation item in the environmental data, Am 2 (i) represents the evaluation value of the i-th evaluation item in the target data, Y i represents the judgment threshold for the i-th evaluation item in the rule information; S5. Repeat steps S2 to S4 until all rule information is processed.

[0009] Furthermore, the multidimensional scoring processor calculates the hallucination index P according to the following formula: ; Among them, m is the number of rule information, k i is the weight of the i-th rule information, Vc(i) is the check value of the i-th rule information; When the hallucination index is greater than the hallucination threshold, it indicates that the target data has the corresponding hallucination type.

[0010] The beneficial effects achieved by the present invention are: This system does not directly detect and analyze the output content of the large model, but first analyzes the input questions of the large model, determines the type of hallucinations that are likely to occur based on the input content type, greatly narrows the analysis scope, and then performs targeted detection and analysis on the output content, which can effectively improve the accuracy of hallucination detection, and make targeted interventions and adjustments based on the detected hallucination types to optimize the output content of the large model. At the same time, this system performs real-time hallucination detection during the output of the large model, replacing the solution of hallucination detection on the complete output content in ordinary systems. It can promptly discover and modify hallucination content, avoid the impact of multiple hallucination contents appearing at the same time on the analysis, and reduce the difficulty of detection.

[0011] To further understand the features and technical contents of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are only for reference and description and are not intended to limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic diagram of the overall structural framework of the present invention; Figure 2 This is a schematic diagram of the input parsing module of the present invention; Figure 3 This is a schematic diagram of the output monitoring module of the present invention; Figure 4 This is a schematic diagram of the intervention and modification module of the present invention; Figure 5 This is a schematic diagram of the rule execution unit of the present invention; Figure 6 This is a comparison chart of the number of hallucinations actually tested in the present invention. DETAILED DESCRIPTION

[0013] The following is an explanation of the embodiments of the present invention through specific embodiments. Those skilled in the art can understand the advantages and effects of the present invention from the contents disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and the details in this specification can also be modified and changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. In addition, the drawings of the present invention are only simple schematic illustrations and are not depicted according to actual sizes. It is stated in advance. The following embodiments will further explain the relevant technical contents of the present invention in detail, but the disclosed contents are not intended to limit the scope of protection of the present invention.

[0014] Embodiment 1: This embodiment provides a system for real-time monitoring and intervention of large models to generate hallucination content, combined with Figure 1 , including large model docking module, input parsing module, output monitoring module and intervention modification module; The large model docking module is used to perform data docking with the large model system, the input parsing module is used to perform content parsing on the input information of the large model, the output monitoring module is used to perform hallucination monitoring on the output content of the large model, and the intervention modification module is used to modify and adjust the hallucination content; The large model docking module includes an input detection unit, an output interception unit and an output feedback unit. The input detection unit is used to detect the input event of the large model and obtain the input content. The output interception unit is used to intercept the output information of the large model. The output feedback unit is used to feed back the actual output information to the large model. The input parsing module includes an information classification unit, an illusion mapping unit and a rule output unit, wherein the information classification unit is used to classify and identify the input information, the illusion mapping unit is used to map the classified information to the illusion type, and the rule output unit is used to send the illusion judgment rule to the output monitoring module; The output monitoring module includes a content management unit, a rule execution unit and a result output unit, wherein the content management unit is used to manage the data content that needs to be hallucinated, the rule execution unit performs judgment analysis on the data content based on the judgment rule, and the result output unit is used to send the judgment result to the intervention modification module; The intervention modification module includes a result diversion unit, an illusion modification unit and a content output unit. The result diversion unit is used to divert the judgment result, the illusion modification unit is used to modify and adjust the illusion content, and the content output unit is used to send the final content information to the output feedback unit.

[0015] The rule output unit includes an illusion rule base, a rule matching processor and a rule packaging processor, wherein the illusion rule base is used to store rule information for judging illusions, the rule matching processor matches corresponding rule information from the illusion rule base based on the illusion type, and the rule packaging processor is used to package all matched rule information and output them; The content management unit includes a detection content register, a detection content update processor and a detection control processor. The detection content register creates a storage area based on the hallucination intervention project to store the output content of the large model. The detection content update processor is used to update the content information in the storage area. The detection control processor organizes the content information in the storage area into an execution data packet for hallucination detection at each update. The execution data packet includes environment data and target data, the environment data is all output content of the hallucination intervention project at the current time point, and the target data is the sentence content newly intercepted by the hallucination intervention project.

[0016] The rule execution unit includes an environment assessment processor, a rule verification processor and a multidimensional scoring processor. The environment assessment processor is used to evaluate and analyze the environment data. The rule verification processor verifies and analyzes the target data based on the rule information and the environment assessment results. The multidimensional scoring processor scores the target data for satisfying the illusion based on the verification results of multiple rules.

[0017] The process of the rule verification processor performing verification analysis on the target data includes the following steps: S1. Obtaining the evaluation results of environmental data and target data; S2, obtaining a rule information; S3, filtering out evaluation items and corresponding evaluation values ​​in the evaluation results based on the rule information; S4. Calculate the check value Vc under the rule information according to the following formula: ; Among them, n is the number of selected evaluation items, Am 1 (i) represents the evaluation value of the i-th evaluation item in the environmental data, Am 2 (i) represents the evaluation value of the i-th evaluation item in the target data, Y i represents the judgment threshold for the i-th evaluation item in the rule information; S5. Repeat steps S2 to S4 until all rule information is processed.

[0018] The multidimensional scoring processor calculates the hallucination index P according to the following formula: ; Among them, m is the number of rule information, k i is the weight of the i-th rule information, Vc(i) is the check value of the i-th rule information; When the hallucination index is greater than the hallucination threshold, it indicates that the target data has the corresponding hallucination type.

[0019] Embodiment 2: This embodiment includes all the contents of Embodiment 1, and provides a system for real-time monitoring and intervening in the generation of hallucination content by a large model, including a large model docking module, an input parsing module, an output monitoring module, and an intervention modification module; The large model docking module is used to perform data docking with the large model system, the input parsing module is used to perform content parsing on the input information of the large model, the output monitoring module is used to perform hallucination monitoring on the output content of the large model, and the intervention modification module is used to modify and adjust the hallucination content; The large model docking module includes an input detection unit, an output interception unit and an output feedback unit. The input detection unit is used to detect the input event of the large model and obtain the input content. The output interception unit is used to intercept the output information of the large model. The output feedback unit is used to feed back the actual output information to the large model. Combination Figure 2 The input parsing module includes an information classification unit, a hallucination mapping unit and a rule output unit, wherein the information classification unit is used to classify and identify the input information, the hallucination mapping unit is used to map the classified information to the hallucination type, and the rule output unit is used to send the hallucination judgment rule to the output monitoring module; Combination Figure 3 The output monitoring module includes a content management unit, a rule execution unit and a result output unit. The content management unit is used to manage the data content that needs to be hallucinated. The rule execution unit performs judgment and analysis on the data content based on the judgment rule. The result output unit is used to send the judgment result to the intervention modification module; Combination Figure 4 The intervention modification module includes a result diversion unit, an illusion modification unit and a content output unit. The result diversion unit is used to divert the judgment result, the illusion modification unit is used to modify and adjust the illusion content, and the content output unit is used to send the final content information to the output feedback unit; The input detection unit includes an event response processor, an input acquisition processor and a project management processor, wherein the event response processor is used to connect to the input port of the large model and respond to input events, the input acquisition processor is used to obtain user input information, and the project management processor is used to establish a new hallucination intervention project and manage project information; The output interception unit includes a sentence judgment processor, an output acquisition processor and a project attribution processor, wherein the sentence judgment processor is used to connect to the output port of the large model and judge the sentence structure of the output content, the output acquisition processor is used to obtain the complete sentence content, and the project attribution processor is used to add the corresponding hallucination intervention project label to the obtained sentence content; The output feedback unit includes an output management processor, a port selection processor and a content feedback processor, wherein the output management processor is used to receive and manage content information, the port selection processor selects a corresponding output port based on an item tag of the content information, and the content feedback processor is used to feed back the content information to the user based on the output port; The information classification unit includes a keyword extraction processor, a feature orientation processor and a classification judgment processor, wherein the keyword extraction processor is used to extract keywords from the input information, the feature orientation processor is used to guide the keywords to the corresponding feature information, and the classification judgment processor determines the classification type of the input information based on the feature information; The hallucination mapping unit includes a hallucination information library, a mapping relationship library and a hallucination output processor, wherein the hallucination information library is used to store hallucination type information, the mapping relationship library is used to store the mapping relationship between each content type and multiple hallucination types, and the hallucination output processor outputs the mapped hallucination type to the rule output unit based on the content type; The rule output unit includes an illusion rule base, a rule matching processor and a rule packaging processor, wherein the illusion rule base is used to store rule information for judging illusions, the rule matching processor matches corresponding rule information from the illusion rule base based on the illusion type, and the rule packaging processor is used to package all matched rule information and output them; The content management unit includes a detection content register, a detection content update processor and a detection control processor. The detection content register creates a storage area based on the hallucination intervention project to store the output content of the large model. The detection content update processor is used to update the content information in the storage area. The detection control processor organizes the content information in the storage area into an execution data packet for hallucination detection at each update. The execution data packet includes environment data and target data, the environment data is the entire output content of the hallucination intervention project at the current time point, and the target data is the newly intercepted sentence content of the hallucination intervention project; It should be noted that the current target data has not been officially output to the user, so the environment data does not include the target data; When the detection content update processor adds the newly intercepted sentence content to the storage area, it is regarded as an update operation; Combination Figure 5 The rule execution unit includes an environment evaluation processor, a rule verification processor and a multi-dimensional scoring processor. The environment evaluation processor is used to evaluate and analyze the environment data. The rule verification processor verifies and analyzes the target data based on the rule information and the environment evaluation result. The multi-dimensional scoring processor scores the target data for satisfying the illusion based on the verification results of multiple rules. The evaluation result of the environmental data is in the form of evaluation items and evaluation values. The environmental evaluation processor saves the evaluation result of the environmental data after the last hallucination detection, and evaluates the target data. When the target data does not have hallucinations, the evaluation result of the target data is merged with the last evaluation result of the environmental data to obtain a new evaluation result of the environmental data. If the target data has hallucinations, the evaluation result of the target data after intervention and modification is merged with the last evaluation result of the environmental data to obtain a new evaluation result of the environmental data. The process of the rule verification processor performing verification analysis on the target data includes the following steps: S1. Obtaining the evaluation results of environmental data and target data; S2, obtaining a rule information; S3, filtering out evaluation items and corresponding evaluation values ​​in the evaluation results based on the rule information; S4. Calculate the check value Vc under the rule information according to the following formula: ; Among them, n is the number of selected evaluation items, Am 1 (i) represents the evaluation value of the i-th evaluation item in the environmental data, Am 2 (i) represents the evaluation value of the i-th evaluation item in the target data, Y i represents the judgment threshold for the i-th evaluation item in the rule information; S5, repeating steps S2 to S4 until all rule information is processed; The multidimensional scoring processor calculates the hallucination index P according to the following formula: ; Among them, m is the number of rule information, k i is the weight of the i-th rule information, Vc(i) is the check value of the i-th rule information; When the hallucination index is greater than the hallucination threshold, it means that the target data has the corresponding hallucination type; The result output unit includes an illusion summarizing processor and an illusion packaging processor, wherein the illusion summarizing processor is used to summarize the existing illusion information, and the illusion packaging processor is used to package the target data and the summarized illusion information into a target data packet; The result diversion unit includes a receiving identification processor and a diversion control processor, wherein the receiving identification processor is used to receive the packaged target data packet and identify whether there is at least one hallucination, and the diversion control processor performs different processing on the target data packet based on the presence of the hallucination; When the target data packet does not contain an illusion, the shunt control processor directly sends the target data in the target data packet to the content output unit, and when the target data packet contains an illusion, the shunt control processor sends the target data packet to the illusion modification unit; The illusion modification unit includes a modification strategy register, a strategy calling processor and an editing and modification processor, wherein the modification strategy register is used to store the modification strategy of each type of illusion, the strategy calling processor calls the corresponding modification strategy based on the illusion type existing in the target data packet, and the editing and modification processor modifies and adjusts the target data based on the modification strategy; The content output unit includes a target data receiving processor and a response transmission processor, wherein the target data receiving processor is used to receive target data, and the response transmission processor is used to perform response processing on the target data; When the target data receiving processor receives the target data from the result splitting unit, the response transmission processor sends a merge signal to the rule execution unit, and when the target data receiving processor receives the target data from the illusion modification unit, the response transmission processor sends the modified target data to the rule execution unit; The response transmission processor also sends the target data to the output feedback unit; After receiving the merge signal, the rule execution unit directly merges the evaluation items. After receiving the target data, the rule execution unit first evaluates the target data and then merges them. The i appearing in the above text is an ordinal number used to indicate a sequence number and has no actual meaning.

[0020] Some of the code for this system is described below: class LargeModelInterface: """ Large model docking module: handles input, output and feedback.

[0021] """ def detect_input_event(self, input_content): """ Input detection unit: detects and obtains input content.

[0022] """ print("Detect input events...") self.input_data = input_content def capture_output(self, output_content): """ Output capture unit: captures large model output information.

[0023] """ print("Intercept output information...") self.output_data = output_content def feedback_output(self, feedback_content): """ Output feedback unit: Feedback the modified content to the big model or user.

[0024] """ print("Feedback final output...") return feedback_content class InputParsingModule: """ Input parsing module: parses input information and generates judgment rules.

[0025] """ def classify_information(self, input_data): """ Information classification unit: classifies and identifies input information.

[0026] """ print("Classification input information...") self.parsed_info = {"type": "query", "content": input_data} def map_to_hallucination(self, parsed_info): """ Hallucination mapping unit: maps categorical information to hallucination types.

[0027] """ print("Mapping illusion type...") hallucination_type = "factual_inaccuracy" # Example type return hallucination_type def generate_rules(self, hallucination_type): """ Rule output unit: generates judgment rules.

[0028] """ print("Generate judgment rules...") self.rules = {"type": hallucination_type, "checks": ["consistency", "accuracy"]} return self.rules class OutputMonitoringModule: """ Output monitoring module: monitors the content generated by the large model.

[0029] """ def __init__(self): self.analysis_result = None def manage_content(self, content): """ Content management unit: manages the data content that needs to be detected.

[0030] """ print("Management content data...") return content def execute_rules(self, content, rules): """ Rule execution unit: judges the content based on the rules.

[0031] """ print("Execution rules...") results = {"content": content, "status": "hallucination_detected"} return results def output_results(self, analysis_results): """ Result output unit: passes the analysis results to the next module.

[0032] """ print("Output analysis results...") self.analysis_result = analysis_results return self.analysis_result class InterventionModule: """ Intervention Modification Module: Processes detected hallucination content.

[0033] """ def __init__(self): self.modified_content = None def distribute_results(self, results): """ Result diversion unit: diverts the judgment results.

[0034] """ print("Diversion processing...") return results["status"] def modify_hallucination(self, content): """ Hallucination Modification Unit: Modify the contents of hallucinations.

[0035] """ print("Modify illusion content...") self.modified_content = content.replace("hallucination_detected", "verified_information") return self.modified_content def output_final_content(self): """ Content output unit: outputs the final revised content.

[0036] """ print("Output final content...") return self.modified_content.

[0037] Now we ask 10 sets of questions to the same large model, and perform manual detection and statistics on the hallucination of the output content when it is connected to this system and when it is not connected to this system, and we get Figure 6 The effect comparison chart is shown.

[0038] The contents disclosed above are only preferred feasible embodiments of the present invention, and do not limit the protection scope of the present invention. Therefore, all equivalent technical changes made using the contents of the present invention specification and drawings are included in the protection scope of the present invention. In addition, the elements therein can be updated as technology develops.

Claims

1. A system for real-time monitoring and intervention of large model-generated hallucination content, characterized in that: It includes large model docking module, input parsing module, output monitoring module and intervention modification module; The large model docking module is used to perform data docking with the large model system, the input parsing module is used to perform content parsing on the input information of the large model, the output monitoring module is used to perform hallucination monitoring on the output content of the large model, and the intervention modification module is used to modify and adjust the hallucination content; The large model docking module includes an input detection unit, an output interception unit and an output feedback unit. The input detection unit is used to detect the input event of the large model and obtain the input content. The output interception unit is used to intercept the output information of the large model. The output feedback unit is used to feed back the actual output information to the large model. The input parsing module includes an information classification unit, an illusion mapping unit and a rule output unit, wherein the information classification unit is used to classify and identify the input information, the illusion mapping unit is used to map the classified information to the illusion type, and the rule output unit is used to send the illusion judgment rule to the output monitoring module; The output monitoring module includes a content management unit, a rule execution unit and a result output unit, wherein the content management unit is used to manage the data content that needs to be hallucinated, the rule execution unit performs judgment analysis on the data content based on the judgment rule, and the result output unit is used to send the judgment result to the intervention modification module; The intervention modification module includes a result diversion unit, an illusion modification unit and a content output unit. The result diversion unit is used to divert the judgment result, the illusion modification unit is used to modify and adjust the illusion content, and the content output unit is used to send the final content information to the output feedback unit.

2. A system for real-time monitoring and intervention of large models to generate hallucination content as claimed in claim 1, characterized in that: The rule output unit includes an illusion rule base, a rule matching processor and a rule packaging processor, wherein the illusion rule base is used to store rule information for judging illusions, the rule matching processor matches corresponding rule information from the illusion rule base based on the illusion type, and the rule packaging processor is used to package all matched rule information and output them; The content management unit includes a detection content register, a detection content update processor and a detection control processor. The detection content register creates a storage area based on the hallucination intervention project to store the output content of the large model. The detection content update processor is used to update the content information in the storage area. The detection control processor organizes the content information in the storage area into an execution data packet for hallucination detection at each update. The execution data packet includes environment data and target data, the environment data is all output content of the hallucination intervention project at the current time point, and the target data is the sentence content newly intercepted by the hallucination intervention project.

3. A system for real-time monitoring and intervention of large models to generate hallucination content as claimed in claim 2, characterized in that: The rule execution unit includes an environment assessment processor, a rule verification processor and a multidimensional scoring processor. The environment assessment processor is used to evaluate and analyze the environment data. The rule verification processor verifies and analyzes the target data based on the rule information and the environment assessment results. The multidimensional scoring processor scores the target data for satisfying the illusion based on the verification results of multiple rules.

4. A system for real-time monitoring and intervention of large models to generate hallucination content as claimed in claim 3, characterized in that: The process of the rule verification processor performing verification analysis on the target data includes the following steps: S1. Obtaining the evaluation results of environmental data and target data; S2, obtaining a rule information; S3, filtering out evaluation items and corresponding evaluation values ​​in the evaluation results based on the rule information; S4. Calculate the check value Vc under the rule information according to the following formula: ； Where n is the number of selected evaluation items, Am1(i) represents the evaluation value of the i-th evaluation item in the environmental data, Am2(i) represents the evaluation value of the i-th evaluation item in the target data, and Y i represents the judgment threshold for the i-th evaluation item in the rule information; S5. Repeat steps S2 to S4 until all rule information is processed.

5. A system for real-time monitoring and intervention of large models to generate hallucination content as claimed in claim 4, characterized in that: The multidimensional scoring processor calculates the hallucination index P according to the following formula: ； Among them, m is the number of rule information, k i is the weight of the i-th rule information, Vc(i) is the check value of the i-th rule information; When the hallucination index is greater than the hallucination threshold, it indicates that the target data has the corresponding hallucination type.

Images