SAT solver optimization system based on large language model

Abstract

The invention discloses an SAT solver optimization system based on a large language model, belongs to the technical field of SAT solver optimization systems, is used for optimizing the performance of an original solver and improving the efficiency of processing a complex SAT problem, and comprises a heuristic function extraction module used for extracting a heuristic function in the original SAT solver; the optimization function selection module is used for selecting the optimized heuristic function in the step; the analysis optimization module is used for communicating with the LLM to obtain an optimized function; the code integration module is used for replacing a corresponding heuristic function in the original SAT solver and constructing a new SAT solver; and the evaluation test module is used for evaluating a new SAT solver and feeding back a result to the optimization function selection module. Compared with the prior art, the advantages of LLM in the aspect of code generation are fully utilized, the corresponding heuristic function can be efficiently and intelligently generated, the performance of the SAT solver is optimized, and the solving efficiency of the SAT solver for different problems is improved.

Description

Technical Field

[0001] The present invention relates to a SAT solver optimization system, in particular to a SAT solver optimization system based on a large language model, and belongs to the technical field of SAT solver optimization systems. Background Art

[0002] On a theoretical level, many combinatorial problems can be expressed in the form of SAT problems and solved by SAT solvers, such as graph coloring problems and vertex cover problems. The study of SAT problems helps to reveal the essence of NP problems and promote the development of computational complexity theory. Finding an efficient SAT solver means that we have more powerful tools when dealing with computationally intensive problems, which will have a profound impact on algorithm design, computational theory, and applied computer science. On a practical level, the improvement and optimization of SAT solvers can directly improve the performance of practical applications. For example, in electronic design automation, SAT solvers are used to verify the correctness of circuit designs. In software engineering, SAT solvers are used to detect errors and vulnerabilities in code. By improving the efficiency of SAT solvers, computing time and resource consumption can be significantly reduced, thereby improving the performance and reliability of the overall system. Therefore, from the perspective of both theoretical research and practical application, it is very necessary to conduct in-depth research on the solution methods of SAT problems in order to obtain acceptable computational complexity. Although the prior art discloses a SAT solver in the comparative document CN109146077B, which includes a bidirectional transfer chain module for bidirectional transfer, A variable to be solved; an implication circuit module, used to provide an implication value for the variable to be solved; a conflict detection module, used to detect whether the current assignment of the assignment module will cause a conflict; an assignment module, used to assign a value to the variable to be solved, and its output value is determined by the current state of the state machine of the variable to be solved and the input signal of the assignment input. The solution of this invention can solve the problem of low processing efficiency of the SAT solver based on software in the existing technology, and achieve the effect of improving processing efficiency, but it cannot realize the functions of cross-domain fusion and adaptive adjustment. For this purpose, a SAT solver optimization system based on a large language model is designed to solve the above problems. Summary of the Invention

[0003] The main purpose of the present invention is to provide a SAT solver optimization system based on a large language model.

[0004] The purpose of the present invention can be achieved by adopting the following technical solutions:

[0005] A SAT solver optimization system based on a large language model, comprising a heuristic function extraction module, an optimization function selection module, an analysis and optimization module, an evaluation and testing module, and a code integration module;

[0006] The heuristic function extraction module is coupled with the optimization function selection module, and the optimization function selection module is coupled with the analysis and optimization module and receives the evaluation and testing module information;

[0007] The analysis and optimization module is coupled with the code integration module, and the code integration module is coupled with the evaluation and testing module.

[0008] Preferably, a heuristic function extraction module is used to extract the heuristic function in the original SAT solver;

[0009] The optimization function selection module is used to select the heuristic function to be optimized in this step;

[0010] Analysis and optimization module, used to communicate with LLM and obtain optimized functions;

[0011] The code integration module is used to replace the corresponding heuristic function in the original SAT solver to construct a new SAT solver;

[0012] The evaluation and testing module is used to evaluate the new SAT solver and feed the results back to the optimization function selection module.

[0013] Preferably, the heuristic function extraction module obtains user input, combines feedback from the LLM, extracts the modified heuristic function, and passes it to the optimization function selection module;

[0014] The user input includes the original SAT solver code and the prompt words provided by the user.

[0015] Preferably, the optimization function selection module selects the heuristic function to be optimized in this step according to the defined heuristic function selection strategy, and passes it to the analysis and optimization module;

[0016] The step refers to a communication with the LLM during the optimization process, and the total number of steps needs to be set during the entire optimization process;

[0017] The heuristic function selection strategy needs to be combined with user description, current optimization stage, and previous optimization result information to make a comprehensive judgment.

[0018] Preferably, the analysis and optimization module obtains a response by providing a prompt word to the LLM, and passes the response to the code integration module;

[0019] The prompt words include the heuristic function to be optimized, the user's description of the function, and the user's description of the LLM task;

[0020] The reply includes the optimized heuristic function and LLM's description of the new heuristic function.

[0021] Preferably, the code integration module processes the response of the LLM, extracts the optimized heuristic function therefrom, replaces the corresponding heuristic function of the original SAT solver in the source code, constructs a new SAT solver, and hands it over to the evaluation and testing module for testing.

[0022] Preferably, the evaluation test module compiles and executes the optimized SAT solver, evaluates its performance, and feeds the results back to the optimization function selection module.

[0023] Beneficial technical effects of the present invention:

[0024] The present invention provides a SAT solver optimization system based on a large language model. The present invention innovatively introduces the large language model (LLM) in the field of natural language processing into the field of SAT solving, cross-domain integration of the latest technologies from different fields, and explores new applications of LLM in logical reasoning and combinatorial optimization problems.

[0025] Different from the traditional end-to-end method, the present invention uses LLM to dynamically generate and optimize heuristic functions, so that the SAT solver can adaptively adjust the strategy according to the specific characteristics of the problem and the feedback information during the solution process, thereby improving the solution efficiency.

[0026] By designing special training and optimization methods, the present invention enables LLM to efficiently learn and generate heuristic functions suitable for SAT problems of different scales and complexities, thereby improving the generalization ability and solving performance of the solver. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a structural diagram of a preferred embodiment of a SAT solver optimization system based on a large language model according to the present invention.

[0028] Figure 2 It is a functional diagram of a preferred embodiment of a SAT solver optimization system based on a large language model according to the present invention.

[0029] Figure 3 It is a functional diagram of a preferred embodiment of a SAT solver optimization system based on a large language model according to the present invention. DETAILED DESCRIPTION

[0030] In order to make the technical solution of the present invention more clear and specific to those skilled in the art, the present invention is further described in detail below with reference to embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0031] Embodiment 1:

[0032] A SAT solver optimization system based on a large language model, such as Figure 1As shown, it includes a heuristic function extraction module 1, an optimization function selection module 2, an analysis and optimization module 3, a code integration module 4, and an evaluation and testing module 5.

[0033] Among them, the heuristic function extraction module 1 is used to extract the heuristic function in the original SAT solver; the optimization function selection module 2 is used to select the heuristic function to be optimized in this step; the analysis and optimization module 3 is used to communicate with the LLM to obtain the optimized function; the code integration module 4 is used to replace the corresponding heuristic function in the original SAT solver and construct a new SAT solver; the evaluation and testing module 5 is used to evaluate the new SAT solver and feed the results back to the optimization function selection module.

[0034] In this embodiment, the heuristic function extraction module 1 obtains user input, combines it with LLM feedback, extracts heuristic functions that may be modified, and passes them to the optimization function selection module 2. The user input includes the original SAT solver code, user-provided prompt words, etc.

[0035] In this embodiment, optimization function selection module 2 selects the heuristic function to be optimized in this step based on a defined heuristic function selection strategy and passes it to analysis and optimization module 3. A step refers to a communication with the LLM during the optimization process, and the total number of steps in the entire optimization process must be set. The heuristic function selection strategy requires a comprehensive judgment based on information such as user description, the current optimization stage, and previous optimization results.

[0036] In this embodiment, Figure 2 As shown, the analysis and optimization module 3 obtains a response by providing a prompt word to the LLM, and passes the response to the code integration module 4. The prompt word includes the heuristic function to be optimized, the user's description of the function, the user's description of the LLM task, etc.; the response includes the optimized heuristic function, the LLM's description of the new heuristic function, etc.

[0037] In this embodiment, Figure 3 As shown, the code integration module 4 processes the LLM's reply, extracts the optimized heuristic function from it, and replaces the corresponding heuristic function of the original SAT solver in the source code, constructs a new SAT solver, and hands it over to the evaluation and testing module 5 for testing.

[0038] In this embodiment, the evaluation and testing module 5 compiles and executes the optimized SAT solver, evaluates its performance, and feeds the results back to the optimization function selection module 2 in preparation for the next optimization.

[0039] The above is only a further embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes based on the technical solutions and concepts of the present invention within the scope disclosed by the present invention, which fall within the scope of protection of the present invention.

Claims

1. A SAT solver optimization system based on a large language model, characterized in that: It includes heuristic function extraction module, optimization function selection module, analysis and optimization module, evaluation and testing module and code integration module; The heuristic function extraction module is coupled with the optimization function selection module, and the optimization function selection module is coupled with the analysis and optimization module and receives the evaluation and testing module information; The analysis and optimization module is coupled with the code integration module, and the code integration module is coupled with the evaluation and testing module.

2. The SAT solver optimization system based on a large language model according to claim 1, characterized in that: Heuristic function extraction module, used to extract the heuristic function in the original SAT solver; An optimization function selection module is used to select the heuristic function to be optimized in this step; Analysis and optimization module, used to communicate with LLM and obtain optimized functions; The code integration module is used to replace the corresponding heuristic function in the original SAT solver to construct a new SAT solver; The evaluation test module is used to evaluate the new SAT solver and feed the result back to the optimization function selection module.

3. The SAT solver optimization system based on a large language model according to claim 2, characterized in that: The heuristic function extraction module obtains user input, combines feedback from the LLM, extracts the modified heuristic function, and passes it to the optimization function selection module; The user input includes the original SAT solver code and the prompt words provided by the user.

4. The SAT solver optimization system based on a large language model according to claim 3, characterized in that: The optimization function selection module selects the heuristic function to be optimized in this step according to the defined heuristic function selection strategy, and passes it to the analysis and optimization module; The step refers to a communication with LLM during the optimization process, and the total number of steps needs to be set during the entire optimization process; The heuristic function selection strategy needs to be combined with user description, current optimization stage, and previous optimization result information to make a comprehensive judgment.

5. The SAT solver optimization system based on a large language model according to claim 4, characterized in that: The analysis and optimization module obtains a response by providing a prompt word to the LLM, and passes the response to the code integration module; The prompt words include the heuristic function to be optimized, the user's description of the function, and the user's description of the LLM task; The reply includes the optimized heuristic function and LLM's description of the new heuristic function.

6. The SAT solver optimization system based on a large language model according to claim 5, characterized in that: The code integration module processes the reply of LLM, extracts the optimized heuristic function from it, replaces the corresponding heuristic function of the original SAT solver in the source code, constructs a new SAT solver, and hands it over to the evaluation and testing module for testing.

7. The SAT solver optimization system based on a large language model according to claim 6, characterized in that: The evaluation test module compiles and executes the optimized SAT solver, evaluates its performance, and feeds the results back to the optimization function selection module.

Images