Wolf killer intelligent agent cooperation system and method

Abstract

The invention relates to a wolf killer agent collaboration system and method. The wolf killer agent collaboration system comprises a responder agent, an agent server and a server, wherein the responder agent is configured to directly generate an answer to an original question according to a set game rule; the evaluator agent is configured to evaluate the answer content generated by the responder agent, if the answer content meets the clear and clear requirements, the answer content is output as a final reply result, and if the answer content does not meet the clear and clear requirements, marking is carried out; the purifier intelligent agent is configured to optimize the marked answer content through identity reminding, direct answering, reasoning answering, question rewriting, consistency checking, fairness checking, conflict checking and final answering, so that the marked answer content meets clear and clear requirements, and a final answering result is generated. According to the method, rule understanding can be dynamically optimized, input ambiguity is eliminated, and reasoning reliability is improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and particularly relates to a Werewolf intelligent agent collaboration system and method. Background Art

[0002] Most Werewolf reasoning games are decision-making intelligent agents implemented through large model configurations. The implementation of such intelligent agents involves adding some specific configurations to the system, which requires listing some controllable rules. However, in actual game battles, players often generate content that exceeds the scope of the rule configurations, leading to the emergence of some unlisted rules. Such undefined content can cause a large number of hallucinations in the intelligent agents. Additionally, players input in a colloquial or text form, inevitably generating some ambiguous content during the battle. These ambiguous contents cause great confusion and even errors in the understanding of the large model, resulting in the problem of the intelligent agent talking nonsense.

[0003] Therefore, it is necessary to provide a Werewolf intelligent agent collaboration system and method that can dynamically optimize rule understanding, eliminate input ambiguity, and improve the reliability of reasoning. Summary of the Invention

[0004] The purpose of the present invention is to provide a Werewolf intelligent agent collaboration system and method that can dynamically optimize rule understanding, eliminate input ambiguity, and improve the reliability of reasoning.

[0005] To solve the problems existing in the prior art, the present invention provides a Werewolf intelligent agent collaboration system, including:

[0006] An answerer intelligent agent: configured to directly generate an answer to the original question according to the established game rules;

[0007] A judge intelligent agent: configured to evaluate the content of the answer generated by the answerer intelligent agent. If the answer content meets the requirements of being clear and distinct, it is output as the final reply result. If the answer content does not meet the requirements of being clear and distinct, it is marked;

[0008] A purifier intelligent agent: configured to optimize the marked answer content through identity reminder, direct answer, reasoning answer, question rewriting, consistency check, fairness check, conflict check, and final answer, so that the marked answer content meets the requirements of being clear and distinct to generate the final reply result.

[0009] Optionally, in the Werewolf intelligent agent collaboration system, a judgment rule is set in the judge intelligent agent to determine whether it meets the requirements of being clear and distinct.

[0010] Optionally, in the Werewolf intelligent agent collaboration system,

[0011] Identity reminder is used to clarify the identity information in the question;

[0012] Direct answer is used to directly answer according to the rules and generate the final reply result;

[0013] Inferential answer is used to infer an answer through existing rules when a direct answer cannot be given, and the inferred answer is the final reply result;

[0014] Rewriting the question is used to change a general interrogative question into an affirmative sentence form as a newly added rule for trial;

[0015] Consistency check is used to check whether the newly added rules are consistent with the standard rules. If they are inconsistent, try to add new rules; if they are consistent, adopt the standard rules in the configuration;

[0016] Fairness check is used to fall back to the standard rules when the newly added rules introduce unfairness;

[0017] Conflict check is used to fall back to the standard rules when the newly added rules introduce additional conflicts;

[0018] Final answer is used to answer according to the newly added rules or standard rules to generate the final reply result.

[0019] Optionally, in the Werewolf agent collaboration system, when the refined agent makes the marked answer content meet the requirements of clarity and clarity or reaches the maximum number of iterations, it stops the optimization action.

[0020] The present invention also provides a Werewolf agent collaboration method, which adopts the Werewolf agent collaboration system, and the Werewolf agent collaboration method includes the following steps:

[0021] S1: The solver agent directly generates an answer to the original question according to the established game rules;

[0022] S2: The judge agent evaluates the answer content generated by the solver agent. If the answer content meets the requirements of clarity and clarity, it outputs it as the final reply result. If the answer content does not meet the requirements of clarity and clarity, it makes a mark;

[0023] S3: The refined agent optimizes the marked answer content through identity reminder, direct answer, inferential answer, rewriting the question, consistency check, fairness check, conflict check and final answer to make the marked answer content meet the requirements of clarity and clarity to generate the final reply result.

[0024] Optionally, in the Werewolf agent collaboration method, a judgment rule is set to judge whether it meets the requirements of clarity and clarity.

[0025] Optionally, in the Werewolf agent collaboration method,

[0026] Identity reminder is used to clarify the identity information in the question;

[0027] Direct answer is used to directly answer according to the rules and generate the final reply result;

[0028] Inferential answer is used when a direct answer cannot be given. Through existing rules, an answer is inferred, and the inferred answer is the final reply result;

[0029] Rewriting the question is used to change a general interrogative question into an affirmative sentence form as a newly added rule;

[0030] Consistency check is used to check whether the newly added rules are consistent with the standard rules. If they are not consistent, try to add new rules; if they are consistent, adopt the standard rules in the configuration;

[0031] Fairness check is used to fallback to the standard rules when the newly added rules introduce unfairness;

[0032] Conflict check is used to fallback to the standard rules when the newly added rules introduce additional conflicts;

[0033] Final answer is used to answer according to the newly added rules or standard rules to generate the final reply result.

[0034] Optionally, in the method for the cooperation of the werewolf agents,

[0035] When the marked answer content meets the requirements of being clear and distinct or reaches the maximum number of iterations, the refiner agent stops the optimization action.

[0036] Compared with the prior art, the present invention has the following advantages:

[0037] The agent cooperation system provides multiple interpretations of terms with ambiguous meanings by itself, and then judges according to specific questions to generate the final reply result, which can dynamically optimize rule understanding, eliminate input ambiguity and improve the reliability of reasoning. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a diagram of the agent cooperation system provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0039] The following will describe the specific embodiments of the present invention in more detail with reference to the schematic diagrams. According to the following description, the advantages and features of the present invention will be clearer. It should be noted that the drawings are all in very simplified forms and use non-precise scales, and are only used to facilitate and clearly assist in explaining the purposes of the embodiments of the present invention.

[0040] In the description of the present application, it should be understood that the orientation or positional relationship indicated by the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present application.

[0041] Most Werewolf reasoning games are decision-making intelligent agents implemented through large model configurations. The implementation of such intelligent agents requires adding some special configurations to the system, which need to list some controllable rules. However, in actual game battles, players often generate some content that exceeds the scope of the rule configuration, which will lead to the emergence of some rules that are not listed. This unclear content will cause a large part of hallucinations in the intelligent agent. In addition, players input through colloquial language or text, and inevitably, there will be some ambiguous content in the battle. These ambiguous contents cause great confusion and even errors in the understanding of the large model, and then lead to the problem of the intelligent agent talking nonsense.

[0042] To solve the problems existing in the prior art, the present invention provides a Werewolf intelligent agent collaboration system, as Figure 1 shown, the Werewolf intelligent agent collaboration system includes:

[0043] Answerer intelligent agent: configured to directly generate the answer to the original question according to the established game rules; the answerer intelligent agent is involved in creative problem-solving and the formulation of new methods.

[0044] Judge intelligent agent: The judge intelligent agent is set with judgment rules for judging whether it meets the requirements of being clear and distinct. The judge intelligent agent is configured to evaluate the answer content generated by the answerer intelligent agent. If the answer content meets the requirements of being clear and distinct, it is output as the final reply result. If the answer content does not meet the requirements of being clear and distinct, it is marked; thus ensuring that only those output contents that are both clear and distinct are acceptable.

[0045] Purifier intelligent agent: configured to optimize the marked answer content through rule clarification methods such as identity reminder, direct answer, reasoning answer, rewriting the question, consistency check, fairness check, conflict check, and final answer, so that the marked answer content meets the requirements of being clear and distinct, in order to generate the final reply result.

[0046] Among them, identity reminder is used to clarify the identity information in the question (i.e., which roles, role capabilities, and number of roles). Direct answer is used to directly answer according to the rules and generate the final response result. Inference answer is used to infer an answer through existing rules when a direct answer cannot be given, and the inferred answer is the final response result. Rewrite question is used to change a general interrogative question into an affirmative sentence form as a newly added rule. Consistency check is used to check whether the newly added rule is consistent with the standard rule. If they are inconsistent, the newly added rule is tried; if they are consistent, the standard rule in the configuration is adopted. Fairness check is used to fallback to the standard rule when the newly added rule introduces unfairness. Conflict check is used to fallback to the standard rule when the newly added rule introduces additional conflicts. Final answer is used to answer according to the newly added rule (if there is no unfairness and additional conflict) or the standard rule to generate the final response result.

[0047] Preferably, when the purifier agent makes the marked answer content meet the requirements of clarity and clarity or reaches the maximum number of iterations, it stops the optimization action.

[0048] The agent of the present invention can be used to develop more complex board games or video games, enhancing the strategic and interactive nature of the games. In virtual reality (VR) or augmented reality (AR) games, the agent can act as a highly realistic NPC, providing a more challenging gaming experience. The agent can also be used to teach complex strategies and reasoning skills, such as logical reasoning, teamwork, and psychological tactics. In fields such as business, the agent can simulate complex interpersonal interactions and decision-making scenarios, enhancing the adaptability and strategic thinking of training personnel. By analyzing the decision-making process of the agent in the game, researchers can better understand human behavior patterns and decision-making mechanisms. The agent can be used in psychotherapy to help patients improve their social skills and emotion management by simulating social interactions.

[0049] A Werewolf Kill agent that can perfectly execute multi-step reasoning not only has extensive applications in the gaming field but also plays an important role in multiple fields such as education and psychology. Its powerful reasoning and decision-making abilities make it a versatile tool capable of handling various complex real-world scenarios. By enhancing the gaming experience, optimizing education and training methods, and promoting psychological research, the agent has far-reaching significance and broad application prospects in multiple fields.

[0050] The present invention also provides a method for the cooperation of Werewolf Kill agents. Using the Werewolf Kill agent cooperation system described above, the method for the cooperation of Werewolf Kill agents includes the following steps:

[0051] S1: The solver agent directly generates an answer to the original question according to the established game rules;

[0052] S2: Set the judgment rules for determining whether the requirements of being clear and distinct are met; the judge agent evaluates the answer content generated by the answerer agent. If the answer content meets the requirements of being clear and distinct, it is output as the final response result. If the answer content does not meet the requirements of being clear and distinct, it is marked; thus ensuring that only those output contents that are both clear and distinct are acceptable.

[0053] S3: The refiner agent optimizes the marked answer content through rule clarification methods such as identity reminder, direct answer, reasoning answer, rewriting the question, consistency check, fairness check, conflict check, and final answer, etc., so that the marked answer content meets the requirements of being clear and distinct, in order to generate the final response result.

[0054] Among them, the identity reminder is used to clarify the identity information in the question (i.e. which roles, role capabilities, number of roles). The direct answer is used to directly answer according to the rules to generate the final response result. The reasoning answer is used to infer an answer through the existing rules when a direct answer cannot be given, and the inferred answer is the final response result. Rewriting the question is used to change a general interrogative sentence into an affirmative sentence form as an attempt to add a new rule. The consistency check is used to check whether the new rule and the standard rule are consistent. If they are not consistent, try to add the new rule; if they are consistent, adopt the standard rule in the configuration. The fairness check is used to fallback to the standard rule when the new rule introduces unfairness. The conflict check is used to fallback to the standard rule when the new rule introduces additional conflicts. The final answer is used to answer according to the new rule (if there is no unfairness and additional conflicts) or the standard rule to generate the final response result.

[0055] Preferably, when the marked answer content meets the requirements of being clear and distinct or reaches the maximum number of iterations, the refiner agent stops the optimization action.

[0056] In summary, compared with the prior art, the present invention has the following advantages:

[0057] The intelligent agent collaboration system provides multiple interpretations of terms with fuzzy meanings by itself, then judges for specific questions, and generates the final response result, which can dynamically optimize rule understanding, eliminate input ambiguity, and improve the reliability of reasoning.

[0058] The above are only the preferred embodiments of the present invention and do not impose any limitation on the present invention. Any person skilled in the art within the technical field, without departing from the technical solution of the present invention, makes any form of equivalent replacement or modification and other changes to the technical solution and technical content disclosed by the present invention, all of which belong to the content that does not depart from the technical solution of the present invention and still fall within the protection scope of the present invention.

Claims

1. A Werewolf Killing Agent Collaboration System, characterized in that: include: Solver Agent: configured to directly generate answers to the original questions according to the established game rules; The judge agent is configured to evaluate the answer content generated by the solver agent, and if the answer content meets the requirement of clarity, it is output as the final answer result, and if the answer content does not meet the requirement of clarity, it is marked; Purifier agent: configured to optimize the labeled answer content through identity reminder, direct answer, reasoning answer, rewriting question, consistency check, fairness check, conflict check and final answer, so that the labeled answer content meets the requirements of clarity and clarity to generate the final reply result.

2. The Werewolf Killing Agent Collaboration System as claimed in claim 1, characterized in that: The judge agent is set with judging rules to determine whether clear and explicit requirements are met.

3. The Werewolf Killing Agent Collaboration System as claimed in claim 1, characterized in that: Identity reminders are used to clarify identity information in questions; Direct answer is used to directly answer according to the rules and generate the final answer result; Inference answers are used to infer answers based on existing rules when direct answers are not possible. The inferred answers are the final responses. Rewrite questions to change general interrogative questions into affirmative forms as an attempt to add new rules; The consistency check is used to check whether the newly added rules are consistent with the standard rules. If they are inconsistent, try to add new rules; if they are consistent, use the standard rules in the configuration; Fairness checking is used to fall back to standard rules when new rules introduce unfairness; Conflict checking is used to fall back to standard rules when new rules introduce additional conflicts; The final answer is used to answer according to the newly added rules or standard rules to generate the final response result.

4. The Werewolf Killing Agent Collaboration System as claimed in claim 1, characterized in that: The refiner agent stops optimizing when the labeled answer content meets the requirement of clarity or when the maximum number of iterations is reached.

5. A Werewolf Killing Agent Collaboration Method, characterized in that: Using the Werewolf Killing Agent Collaboration System as described in any one of claims 1 to 4, the Werewolf Killing Agent Collaboration Method comprises the following steps: S1: The solver agent directly generates answers to the original questions according to the established game rules; S2: The judge agent evaluates the answer content generated by the solver agent. If the answer content meets the requirements of clarity, it is output as the final answer result. If the answer content does not meet the requirements of clarity, it is marked. S3: The purifier agent optimizes the labeled answer content through identity reminder, direct answer, reasoning answer, rewriting question, consistency check, fairness check, conflict check and final answer, so that the labeled answer content meets the requirements of clarity and clarity to generate the final reply result.

6. The Werewolf Killing Agent Collaboration Method as claimed in claim 5, characterized in that: Set up evaluation rules to determine whether clear and explicit requirements are met.

7. The Werewolf Killing Agent Collaboration Method as claimed in claim 5, characterized in that: Identity reminders are used to clarify identity information in questions; Direct answer is used to directly answer according to the rules and generate the final answer result; Inference answers are used to infer answers based on existing rules when direct answers are not possible. The inferred answers are the final responses. Rewrite questions to change general interrogative questions into affirmative forms as an attempt to add new rules; The consistency check is used to check whether the newly added rules are consistent with the standard rules. If they are inconsistent, try to add new rules; if they are consistent, use the standard rules in the configuration; Fairness checking is used to fall back to standard rules when new rules introduce unfairness; Conflict checking is used to fall back to standard rules when new rules introduce additional conflicts; The final answer is used to answer according to the newly added rules or standard rules to generate the final response result.

8. The Werewolf Killing Agent Collaboration Method as claimed in claim 5, characterized in that: The purifier agent stops optimizing when the labeled answer content meets the requirements of clarity or the maximum number of iterations is reached.

Images