39 results about "Modula" patented technology

In some embodiments, a secure chatbot includes a generative language module (GLM) and a prompt manager configured to dynamically update GLM input prompts according to outputs of the GLM. The input prompts instruct the GLM to perform a task (e.g., determine whether a target message received by a user indicates online fraud) and, in response, output a particular token. In response to detecting the token within the output of the GLM, the prompt manager can selectively execute a code snippet identified according to the respective token. Executing the code snippet effectively updates the prompt, which is then fed back to the GLM for a next inference step. In contrast to conventional LLM prompts, the updates to the prompts herein include modifications (e.g., insertion of other tokens, etc.) rather than merely appending inferred tokens to previous prompts. Moreover, the prompt modifications are executed inline, i.e., without submitting a wholly new prompt to the GLM.

The application relates to the technical field of code automatic generation, in particular to a system and method for automatic code generation, which comprises the following steps: a requirement analysis module receives user requirements through multi-modal input, and extracts formatted requirement information by using a BERT language model; a template selection module selects a code template from a plug-in template library; a parameter extraction module extracts and verifies key parameters by using a machine learning algorithm; a template filling module fills the parameters into the template and generates a code framework; a framework verification module verifies and corrects the code framework; a code generation module outputs the corrected code framework as a code file and integrates the code file into a development environment of a user project; and a feedback iteration module receives user feedback, analyzes and optimizes system performance. Through full-process automation and intelligent design, the application greatly improves the efficiency and accuracy of code generation, optimizes user operation experience, and can be widely applied to various programming languages and various development scenarios.

A method may include: obtaining an input query; generating, by using a large language model (LLM), based on the input query, a plan query comprising one or more blank queries related to a first sub-module and one or more operations to be performed by the first sub-module; transmitting, to the first sub-module, (a) a request for generating a first response comprising one or more values corresponding to the one or more blank queries, (b) at least a part of the plan query, and (c) a first guide query, the first guide query being based on a part of the input query associated with the first sub-module; receiving, from the first sub-module, the first response generated by using a sub-language model (LM) of the first sub-module; and generating, by using the LLM, a final response to the input query, based on the first response.

This invention relates to a method for replacing the original exception module with a custom-written timer module using the ST language of CodeSys, belonging to the field of vehicle leveling and servo control technology. A custom TON is designed based on the original TON interface and then called, including the following steps: If not enabled, the timer's delay time is cleared, the internal timing variable is loaded with the delay time count, and the process returns to the next loop; if the internal timing variable is not loaded with the delay time count, it is loaded only once, and timing begins. If the delay time has not yet expired, the delay time is cleared; otherwise, the delay time is set to expire. This method is simple and easy to implement, requiring only the addition of a custom-written timer module and its calling program, without modifying other programs, thus solving the problem of unreliable delay triggering when using the original TON.

This invention discloses an FPGA netlist synthesis method and system based on multi-tool collaboration. The method includes: acquiring the parent layer design code of the FPGA to be synthesized; identifying logic modules in the parent layer design code that cannot be synthesized using a first synthesis tool, as these modules to be split; separating the modules to be split into a second design part, and using the remaining part of the parent layer design code as the first design part, retaining the black-box interface definitions of the modules to be split in situ within the first design part; independently synthesizing the first design part using the first synthesis tool to generate a first netlist; selecting a suitable second synthesis tool to independently synthesize the second design part to generate a second netlist; and filling the entities corresponding to the second netlist into the black-box positions of the first netlist to generate a parent layer netlist file. This method solves the problem of synthesis failure caused by a single synthesis tool's incompatibility with multiple languages ​​and IP sources, and avoids the forced porting of unmodifiable code.

This application provides a method and apparatus for generating code-based module design documents. The method includes: obtaining target code; deleting compiler directives and comments from the target code to obtain code text; identifying and extracting global variables and function structures from the code text; comparing the identifiers in each function structure in parallel with global variables, and comparing the identifiers before the left parenthesis in each function structure in parallel with the function names of other function structures, and associating the global variables and / or function names that match the comparison with the corresponding function structures, wherein the other function structures are the function structures outside the current function structure in each function structure; and generating and outputting a target module design document in a preset format based on each function structure and its associated global variables and / or function names. This achieves efficient and accurate generation of module design documents.

The present application relates to the field of artificial intelligence and computer, more particularly, the present application relates to a language model speculation reasoning acceleration method and system based on step mapping, which is applied to a computing device deployed with a target language model and a draft language model, the method sets the maximum number of steps of draft generation and limits the running step number K, constructs a draft language model containing M identical processing modules, pre-configures and solidifies a static step mapping function g(k)=k, calibrates the single-step cost offline under the reference input shape (,,D) and sets the threshold value, forms a reproducible configuration record, during reasoning, the target hidden state vector h is reused and fused with the token embedding vector e in the fusion layer, and only the corresponding processing module is activated according to the step to generate the candidate sequence; the target language model verifies the candidate sequence vectorization to obtain the accepted prefix length A, and according to the acceptance statistics, the speculation step number K is adaptively adjusted up or down according to the period T, and the update result is written into the configuration record to support subsequent reproducible calling.

This application discloses a method and system for generating code review comments on modified code, belonging to the field of computer technology. The method includes: parsing the submitted source code through an IDEA plugin module to abstract the modified file into an abstract syntax tree; extracting information from the abstract syntax tree to obtain extracted information, which includes: the package name, class name, class comments, signatures, comments, and field declarations of other unmodified methods in the class; storing the context information of the modified code obtained from the extracted information into a database; querying the context information through a backend service module and constructing prompt words based on the context information; and guiding a large language model to generate code review comments through the prompt words. This generation method can automatically generate code review comments, thereby improving the intelligence level of code review.

This invention discloses a cross-language compiler optimization method based on graph neural networks. The method uses a UAST module to map the abstract syntax trees (ASTs) of different programming languages ​​to a shared semantic space. The FUAST module then fuses control flow and data flow information to form a flow-enhanced AST representation. The Sub2Vec module extracts key optimization subgraphs such as control flow, function inlining, and loops, and models their local structure and global relationships. The IRGAT module uses a sparse attention mechanism after structural pruning to aggregate features from the subgraphs, generating graph-enhanced sequence vectors. Finally, the SALT module, based on a lightweight Transformer architecture, embeds IRGAT as a structure-aware component, which processes and outputs optimization decisions through a feedforward network. This invention achieves unified semantic understanding, dynamic information fusion, and key region focusing for multi-language programs, and possesses compiler platform versatility.

The application designs a detection method for the defect of the UNSAFE code segment in the Rust programming language, first, based on the rustc_lint module, whether the expression in the program language is contained in the unsafe block written by the developer is judged by using HIR; the structure of the expression is analyzed by using HIR; then, the unsafe operation is analyzed and identified by using MIR; finally, the safe expression or statement in the unsafe block written by the developer is reported; the data flow analysis technology for the function parameter is proposed, the alias propagation chain of the function parameter is found out; it is the key technology required for judging whether the function or method is marked with the unsafe keyword, and the target function is tested by combining the fuzz test technology to judge whether the function must be marked with the unsafe keyword.

The present disclosure relates to a method for determining improved values of parameters for operating a Large-Language-Model, LLM, comprising: generating chunks of documents and input content dependent on the chunks according to several different modes, wherein the respective mode is specified by a respective set of values of the parameters; generating a prompt for the LLM dependent on the input content and a respective question; providing the prompt as an input to the LLM and receiving a respective provisional answer in response from the LLM; and performing a comparison between the provisional answer and a target answer resulting in a score for the respective question; training a machine-learning module using the sets of values of the parameters and the scores for the questions as training data; and performing a search for the improved values of the parameters using the trained ML-module.

Systems and methods are provided for mitigating third-party code vulnerabilities in AI code generation services. A large language model generates an inference based on a user's code generation request. A middleware layer intercepts the inference before its availability within an AI-to-programming client and parses the inference to identify third-party packages. A verification module checks the identified third-party packages against one or more authenticated package registries. If an unauthenticated third-party package is detected, the middleware layer either edits the inference or modifies the inference by replacing with an identified authenticated third-party package replacement. The system enhances the security of AI-assisted code generation by ensuring the use of authenticated third-party packages.

A method for determining improvement parameter values ​​for running a Large-Scale Language Model (LLM). [Solution] The method includes generating chunks of a document and input content dependent on those chunks according to a plurality of different modes, where each mode is specified by a set of values ​​for a parameter; generating prompts for an LLM depending on the input content and each question; providing the prompts as input to the LLM and receiving each tentative answer as a response from the LLM; and performing a comparison between the tentative answers and the target answers to obtain a score for each question as a result; training a machine learning module using the set of parameter values ​​and the scores for the questions as training data; and using the trained ML module to search for improved values ​​for the parameters.

This invention relates to a method, apparatus, and system for security evaluation of large language models based on dynamic custom logic encoding. It constructs a temporary logic mapping protocol for encoding; after acquiring malicious semantic text to be tested, it encodes the malicious semantic text into a sequence of logical instructions according to the temporary logic mapping protocol, constructs test prompts, inputs them into the large language model under test, and determines whether the large language model under test has been jailbroken based on the output. The apparatus includes an acquisition module, a protocol construction module, an encoding module, a construction module, an output acquisition module, and a judgment module; the system is implemented based on the method. This invention can bypass security protection mechanisms based on blacklists and fixed pattern recognition, achieving a high avoidance rate, breaking through static defense mechanisms, touching security vulnerabilities at the logical reasoning level, possessing high adaptability and scalability, ensuring the long-term effectiveness of the evaluation, accurately locating weak points in the model's security strategy, and performing targeted and refined reinforcement.

The application relates to the technical field of large model module intelligent agent execution control, in particular to a large model intelligent agent autonomous execution method and system based on a logic management module. The system comprises a world state management module, a logic management module and a large model module. The large model module generates a logic target sentence according to a user instruction, the logic management module generates a logic result according to the logic target sentence, and the user instruction result further generated by the large model module is fed back to the user. The system adopts a multi-module structure, the world state management module uses a general programming language to realize core scheduling, dynamic execution and control, the logic management module is responsible for planning and reasoning to generate a logic result without illusion, and the large model module understands the user instruction. In the embodiment of the application, the logic management module has complete formal logic derivation capability, and all operations have formal checkability, so that illusion is eliminated from the mechanism.

The application discloses a dynamic link library reflection method based on a C language preprocessor, and comprises the following steps: S1, acquiring the file name of a dynamic link library to be agented and a function list to be agented, wherein the function list comprises the function name, return type and parameter type of each function; S2, constructing a registerer by using the macro definition of the C language, wherein the input of the registerer is the file name of the dynamic link library, the function name to be agented, the return type and the parameter type; S3, generating an agent function by using the registerer in the preprocessing process of the compiler at the compiling stage; S4, registering the function to be agented in the source code by using the registerer in step S3; and S5, calling the function to be agented in S1 in the business, and compiling and running the software together with the registration code in step S4. The application solves the compatibility problem caused by using different version library modules or using the same library module in different versions of the same product.

The application provides a power engineering cost software cross-platform secondary development method, and relates to the field of software development.The method comprises the following steps: dividing a C++ original business module, preparing a to-be-generated header file, dividing the C++ original business module, preparing the to-be-generated header file, and forming information such as a header file, a link library dependency, and a link library directory of each module.Through the use of a path moving mechanism and an erasing mechanism, the existing C++ code function can be reused, and C++ objects and Python object data can be interchanged.The core of the method is to introduce Python language for secondary development, and the advantage lies in simple syntax, dynamic support, which can greatly improve the development efficiency, reduce the learning threshold, and simplify the cross-platform adaptation process.

The application discloses a general development system of programmable normalized script engine, which comprises a VS plug-in module, a project management module and a script syntax module.The VS plug-in module is used for responding to IDE compilation instructions, displaying compilation information and positioning the text corresponding to the compilation information; the project management module is used for distributing compilation instructions, transferring compilation information and generating intermediate files; and the script syntax module is used for registering keywords, data conversion, data validity check, expression analysis and syntax logic check.The application has the advantages of convenient and fast implementation process, stable operation, safety and reliability, wide application range, simple interface, high operation efficiency, and programmable normalized script engine, and has strong universality.

This invention proposes a ensemble communication compression method for distributed training of large language models, comprising: constructing a unified compression acceleration programming model and defining a set of core interfaces for compression acceleration; configuring a custom compression algorithm for the target training data type based on the unified compression acceleration programming model, and compiling the custom compression algorithm into a dynamically loadable link library file; submitting the environment variables configured for the custom compression algorithm through the configuration registration interface, and initializing the compression module of the compression-accelerated ensemble communication library through the environment variables; executing 3D parallel training of the large language model, and during the training process, the compression-accelerated ensemble communication library calls the corresponding custom compression algorithm to compress or decompress the communication data when performing ensemble communication operations according to the preset mapping relationship between communication primitives and compression algorithms, thereby completing data transmission across GPU nodes. This invention improves communication efficiency.

Devices, systems, and methods for language model prompt engineering include a method of: providing, by at least one first processor, to a language model, a first prompt including a first task and instructing the language model to generate a set of task-specific reasoning modules; generating, by the language model, the set of task-specific reasoning modules based on the first prompt, the set of task-specific reasoning modules including a first reasoning module; providing, by at least one second processor, to the language model, a second prompt including the first task, a second task, and the first reasoning module, wherein the second prompt instructs the language model to generate an initial reasoning structure defining how to respond to the first task without providing a response; and generating, by the language model, the initial reasoning structure based on the second prompt.

This invention relates to the fields of computer software and industrial control technology, and discloses a method for configuring communication module software based on the PROFINET protocol, comprising: S1: using a Qt5-based XML parsing library combined with the GSDML file standard structure to independently parse the PROFINET controller device description file, i.e., the GSDML file, and generate device parsing information; S2: displaying the description item information designed according to the device parsing information and the configuration item information set by the user on the software interface, respectively, for configuring the parameters of the controller module of the software; S3: saving the configuration item information to a specified file format, inputting it into the compiler software for execution, and simultaneously inputting the GSDML file into the compiler software and executing the call to generate a standard configuration file. This invention independently and controllably completes the configuration of the PROFINET slave controller, integrating the configuration of the PROFINET controller into domestic programming software, achieving independent control and ease of use for users.