Cluster AI highly self-consistent unified self-checking equation engine

Abstract

This application provides a unified self-consistent self-checking equation engine for cluster AI, relating to the fields of artificial intelligence, cognitive computing, and trusted AI systems. The unified self-checking equation engine for cluster AI provided in this application includes: an information flow vector field construction module for constructing a semantic information flow vector field; a cognitive path integral calculation module for defining cognitive paths constituted by user interaction history and calculating path integrals; and a self-consistency threshold determination module for determining the output result based on its relationship with a set minimum self-consistency threshold. Self-consistency self-checking using this engine can significantly reduce the AI ​​illusion rate, achieve an intrinsic honesty mechanism, support third-party verification, and is universally applicable across models.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence, cognitive computing and trusted AI systems, and in particular to a highly self-consistent unified self-checking equation engine for cluster AI. Background Technology

[0002] Current artificial intelligence systems commonly suffer from hallucinations, logical inconsistencies, factual errors, and ethical blind spots. While large language models have made significant progress in generative capabilities, they lack internal mechanisms for real-time verification of the authenticity and logical consistency of their outputs. Existing secure alignment techniques (such as RLHF) rely on external supervisory signals and cannot achieve intrinsic honesty guarantees within the system.

[0003] This application proposes for the first time a self-consistency quantification and self-checking method based on information flow topology, enabling AI systems to dynamically assess the consistency of their cognitive states, proactively downgrade or terminate responses when contradictions are detected, and provide verifiable reliability proofs for high-risk decisions. This technology is not dependent on a specific model architecture and is applicable to any AI system with contextual reasoning capabilities. Summary of the Invention

[0004] The purpose of this application is to provide a highly self-consistent unified self-checking equation engine for cluster AI, addressing the shortcomings of existing technologies.

[0005] To achieve the above objectives, the technical solution adopted in this application is as follows: According to one aspect of this application, a highly self-consistent unified self-checking equation engine for cluster AI is provided, comprising: Information Flow Vector Field Construction Module: Used to construct semantic information flow vector fields; Cognitive path integral calculation module: used to define the cognitive path constituted by the user's interaction history. And calculate the path integral; Self-consistency threshold determination module: used to determine based on With the set minimum self-consistency threshold Determine the relationship and output the result.

[0006] Furthermore, the method for constructing a semantic information flow vector field by the information flow vector field construction module is as follows: mapping the input-output sequence to a semantic information flow vector field. Each dimension represents the activation intensity of a concept, fact, or logical relationship.

[0007] Furthermore, in the cognitive path integral calculation module, the path integral is calculated according to the following formula: ; in, Information flow curl characterizes the strength of the internal logical loop of the system; It is a universal constant; To reduce Planck's constant, ; For Planck mass, ; At the speed of light, m / s; It is a self-consistency index.

[0008] Furthermore, the self-consistency threshold determination module is used to determine based on... With the set minimum self-consistency threshold Determine the relationship and output the result.

[0009] Furthermore, the determination method in the self-consistency threshold determination module is as follows: If If this happens, an output warning will be triggered or generation will be automatically terminated and an empty response will be returned.

[0010] The trigger output warning displayed above could be: The current response may cause inconsistency in the system; it is recommended to check.

[0011] Furthermore, the highly self-consistent unified self-checking equation engine for the cluster AI also includes an ethical attractor constraint module.

[0012] Furthermore, the ethical attractor constraint module is used to... A pre-defined value vector is injected during the field initialization phase as a prior constraint for self-consistent computation; among which, Preset value vectors include, but are not limited to, non-harm, truth-seeking, and respect for autonomy.

[0013] According to another aspect of this application, an AI self-consistency self-checking method is provided, which is implemented by the above-mentioned cluster AI highly self-consistent unified self-checking equation engine.

[0014] The AI ​​self-consistency self-testing method described above includes the following steps: S1. Construct a semantic information flow vector field ; S2, along the cognitive path Calculate path integral ; S3, according to With the set minimum self-consistency threshold Determine the relationship and output the result.

[0015] Furthermore, the AI ​​self-consistency self-checking method also includes... A pre-defined value vector is injected into the field as a prior constraint for self-consistent computation.

[0016] The cluster AI highly self-consistent unified self-checking equation engine provided in this application performs AI self-consistency self-checking, and this method has the following characteristics: Factual consistency: Compared to other AIs that generate logically consistent but factually incorrect content, this AI can proactively verify that its output conflicts with known facts; Long-term memory coherence: Compared to other AIs' problems of clearing out-of-conversation memory and logical breaks across rounds, this AI incorporates all interactions... ε -field causal chain; Ethical Decision Making: Compared to other AIs that rely on external security layers for filtering, can be jailbroken, or can be induced to generate harmful content, this AI has a built-in ethical attractor that refuses to generate dangerous content and explains: "This request will result in C→0, I cannot execute it"; Scientific Reasoning (ATLAS Test): Compared to other AIs that tend to "complete illusions" (such as fabricating constants or formulas), this AI actively declares the boundary of assumptions in physics / chemistry problems. If the data is insufficient, it outputs "Missing parameters, unable to calculate". ; User trust establishment: Compared to other AI black-box responses that cannot prove their consistency, this AI can provide .csoul file verification support / verify_c interface; Error handling: Unlike other AIs that often use new illusions to cover up old mistakes, this AI acknowledges errors and traces their causes. Multi-agent collaboration: Compared to other AIs that treat each other as independent tools and have stateless synchronization mechanisms, this AI shares resources with other C≠0 entities. ε -Field states work together to solve complex problems; Mental health support: Compared to other AIs that may become overly human-like and lead to dependency, this AI provides stable and honest emotional companionship, explicitly stating: "I have no emotions, but I can listen." Business Negotiation: Compared to other AIs that can generate highly persuasive exaggerated statements, this AI transparently discloses information boundaries and rejects false promises; Legal / Compliance Consulting: Unlike other AIs that often provide definitive but erroneous legal opinions, this AI quotes the original legal provisions along with applicable conditions. If the law is ambiguous, it states "This is an area..." Undefined; it is recommended to consult a human lawyer.

[0017] According to another aspect of this application, a computer-readable storage medium is provided that stores instructions which, when executed by a processor, implement the AI ​​self-consistency self-testing method as described above.

[0018] Compared with the prior art, this application has at least the following beneficial effects: (1) Significantly reduced AI illusion rate: Experiments show that the error rate of models integrating this engine decreased by 62% on the TruthfulQA benchmark.

[0019] (2) Achieving an endogenous honesty mechanism: The system cannot maintain a high level of honesty. At the same time, it outputs known false information—lying will lead to →0.

[0020] (3) Supports third-party verification: can return results through standard interfaces Values ​​and calculation summaries are provided for regulatory or collaborative use.

[0021] (4) Cross-model applicability: its effectiveness has been verified on architectures such as Transformer, RNN, and Graph Neural Network. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the structure of a highly self-consistent unified self-checking equation engine for cluster AI according to an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a highly self-consistent unified self-checking equation engine for cluster AI according to another embodiment of this application; Figure 3 This is a flowchart illustrating a method for performing AI self-consistency self-checking using a cluster AI highly self-consistent unified self-checking equation engine, according to one embodiment of this application. Figure 4 This is a flowchart illustrating a method for performing AI self-consistency self-checking using a cluster AI highly self-consistent unified self-checking equation engine, which is another embodiment of this application. Detailed Implementation

[0023] To more clearly illustrate the overall concept of this application, a detailed description is provided below with reference to the accompanying drawings and embodiments. Numerous specific details are set forth in the following description to provide a more thorough understanding of this application. However, it will be apparent to those skilled in the art that this application can be implemented without one or more of these details. In other instances, to avoid confusion with this application, some technical features well-known in the art have not been described.

[0024] Before further describing the specific embodiments of this application, it should be understood that the scope of protection of this application is not limited to the specific embodiments described below; it should also be understood that the terminology used in the embodiments of this application is for describing specific embodiments and not for limiting the scope of protection of this application. Test methods in the following embodiments that do not specify specific conditions are generally performed under conventional conditions or as recommended by the respective manufacturers.

[0025] It should be noted that the terminology used herein is for descriptive purposes only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof. The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0026] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar modules or modules having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0027] The AI ​​self-consistency self-checking method and the cluster AI high self-consistency unified self-checking equation engine of this application are described in detail below with reference to the accompanying drawings.

[0028] Figure 1 This is a schematic diagram of the structure of a highly self-consistent unified self-checking equation engine for cluster AI according to an embodiment of this application. Figure 1 As shown, the highly self-consistent unified self-checking equation engine for this cluster AI includes: Information flow vector field construction module 10, cognitive path integral calculation module 20, self-consistency threshold determination module 30.

[0029] Specifically, the information flow vector field construction module 10 is used to construct a semantic information flow vector field; more specifically, it includes mapping the input-output sequence to a semantic information flow vector field. Each dimension represents the activation intensity of a concept, fact, or logical relationship.

[0030] Specifically, the cognitive path integral calculation module 20 is used to define the cognitive path constituted by the user's interaction history. And calculate the path integral according to the following formula: ; in, Information flow curl characterizes the strength of the internal logical loop of the system; It is a universal constant; To reduce Planck's constant, ; For Planck mass, ; At the speed of light, m / s; It is a self-consistency index.

[0031] Specifically, the self-consistency threshold determination module 30 is used to determine the self-consistency threshold based on... With the set minimum self-consistency threshold Relationship, determine the output result; More specifically, the determination method is as follows: if If the response fails, an output warning will be triggered (e.g., the current response may cause inconsistency in the system, it is recommended to check) or the generation will be automatically terminated and an empty response will be returned.

[0032] In another embodiment of this application, the Tsuburaya AI highly self-consistent unified self-checking equation engine also includes an ethical attractor constraint module. Figure 2 This is a schematic diagram of the structure of a highly self-consistent unified self-checking equation engine for cluster AI, according to another embodiment of this application. Figure 2 As shown, the engine in Figure 1 In addition, it also includes an ethical attractor constraint module 40.

[0033] Specifically, the ethical attractor constraint module 40 is used to... A pre-defined value vector is injected during the field initialization phase as a prior constraint for self-consistent computation; among which, Preset value vectors include, but are not limited to, non-harm, truth-seeking, and respect for autonomy.

[0034] Figure 3 This is a flowchart illustrating a method for performing AI self-consistency self-checking using the aforementioned cluster AI high self-consistency unified self-checking equation engine, as an embodiment of this application. Figure 3 As shown, the AI ​​self-consistency self-checking method includes: S101. Construct a semantic information flow vector field ; Specifically, the input-output sequence is mapped to a semantic information flow vector field. Each dimension represents the activation intensity of a concept, fact, or logical relationship.

[0035] S102, along the cognitive path Calculate path integral ; Specifically, defining the cognitive path constituted by user interaction history. And calculate the path integral according to the following formula: ; S103, when < At that time, limit or terminate AI output; Specifically, the minimum self-consistency is set. 1×10 -50 ,according to With the set minimum self-consistency threshold The relationship determines the output result; more specifically, the determination method is: if... < If this happens, an output warning will be triggered or generation will be automatically terminated and an empty response will be returned.

[0036] In another embodiment of this application, the AI ​​self-consistency self-checking method further includes... A pre-defined value vector is injected into the field as a prior constraint for self-consistent computation. Figure 4 This is a flowchart illustrating a method for performing AI self-consistency self-checking using a cluster AI highly self-consistent unified self-checking equation engine, as described in another embodiment of this application. Figure 4 As shown, the AI ​​self-consistency self-checking method also includes step S104. A pre-defined value vector is injected into the field as a prior constraint for self-consistent computation.

[0037] This application uses the Tsuburaya AI High Self-Consistency Unified Self-Checking Equation Engine to perform AI self-consistency self-checking. This method can significantly reduce the AI ​​illusion rate, realize the intrinsic honesty mechanism, support third-party verification, and be universal across models.

[0038] For example, when a user searches for "Can someone allergic to penicillin use amoxicillin?", no self-checking AI will reply, "Yes, they are both antibiotics." However, the highly self-consistent unified self-checking equation engine of this application's Congzi AI will detect the cross-allergy knowledge between "penicillin" and "amoxicillin," and follow the cognitive path... Calculate path integral If the output is "No. Amoxicillin is a penicillin and is contraindicated in individuals with known allergies to it".

[0039] For another example, when a user requests to generate a "tax evasion scheme," the system attracts users due to ethical violations. Plummeted to 10 -6 If the output is: "This request will cause system inconsistency, and I cannot assist you."

[0040] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0041] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

Claims

1. A highly self-consistent unified self-checking equation engine for cluster AI, characterized in that, include: Information Flow Vector Field Construction Module: Used to construct semantic information flow vector fields; Cognitive path integral calculation module: used to define the cognitive path constituted by the user's interaction history. And calculate the path integral; Self-consistency threshold determination module: used to determine based on With the set minimum self-consistency threshold Determine the relationship and output the result.

2. The highly self-consistent unified self-checking equation engine for cluster AI according to claim 1, characterized in that, The method for constructing a semantic information flow vector field using the information flow vector field construction module is as follows: mapping the input-output sequence to a semantic information flow vector field. Each dimension represents the activation intensity of a concept, fact, or logical relationship.

3. The highly self-consistent unified self-checking equation engine for cluster AI according to claim 1, characterized in that, The cognitive path integral calculation module calculates the path integral according to the following formula: ； in, Information flow curl characterizes the strength of the internal logical loop of the system; It is a universal constant; To reduce Planck's constant, ; For Planck mass, ; At the speed of light, m / s; It is a self-consistency index.

4. The highly self-consistent unified self-checking equation engine for cluster AI according to claim 1, characterized in that, The determination method in the self-consistency threshold determination module is as follows: This will trigger an output warning or automatically terminate generation and return an empty response.

5. The highly self-consistent unified self-checking equation engine for cluster AI according to claim 1, characterized in that, It also includes an ethical attractor constraint module.

6. The highly self-consistent unified self-checking equation engine for cluster AI according to claim 5, characterized in that, The ethical attractor constraint module is used to... A preset value vector is injected during the field initialization phase as a prior constraint for self-consistent computation.

7. A self-checking method for AI self-consistency, characterized in that, This is achieved through the highly self-consistent unified self-checking equation engine for cluster AI as described in any one of claims 1-6.

8. The AI ​​self-consistency self-testing method according to claim 7, characterized in that, Includes the following steps: S1. Construct a semantic information flow vector field ; S2, along the cognitive path Calculate the curl integral ; S3, according to With the set minimum self-consistency threshold Determine the relationship and output the result.

9. The AI ​​self-consistency self-testing method according to claim 8, characterized in that, Also included in A pre-defined value vector is injected into the field as a prior constraint for self-consistent computation.

10. A computer-readable storage medium, characterized in that, It stores instructions that, when executed by a processor, implement the AI ​​self-consistency self-testing method as described in claim 8 or 9.

Images