16 results about "Base code" patented technology

The embodiment of the invention relates to a code completion method and device based on a code repository, and the method comprises the steps: firstly, obtaining a plurality of identifiers in a to-be-completed target code statement, and obtaining the first k lines of codes at the position of a cursor as target code snippets; then, on the basis of the identifiers, a plurality of code members meeting the grammar requirement of the position where the cursor is located are retrieved in a code repository; next, searching the positions of the identifiers and the code members appearing in a code repository, extracting code snippets serving as contexts based on the positions, and determining a code use example document based on m code snippets ranked in the top of the first similarity with the target code snippets; finally, inputting the target cue word into a pre-trained code completion model, and indicating the code completion model to generate a completion code; the target prompt word comprises the code use example document and a target code snippet.

A method for incrementally training a base chorer is disclosed. The method includes iteratively initially training the base chorer with exemplars comprising a single oligonucleotide sequence and generating labeled training data using the initially trained base chorer. In operation (i), the base chorer is further trained with exemplars comprising multiple oligonucleotide sequences and labeled training data is generated using the further trained base chorer. Operation (i) is iteratively repeated to further train the base chorer. In an embodiment, during at least one iteration, the complexity of the neural network configuration loaded into the base chorer is increased. In an embodiment, the labeled training data generated during an iteration is used to train the base chorer during a subsequent iteration.

The invention discloses a code library question answering method and device based on a large model, and relates to the technical field of software engineering. The method comprises the steps of performing word segmentation on code files to obtain word units, performing statistics on the number of the word units in each code file, initializing the directory depth of the code files, further determining the hierarchical relationship of the code files to generate a code file set, and inputting the code file set into a large language model to generate a natural language description set; inputting the code file set and the natural language description set into a large language model to obtain code interpretation; segmenting the code explanation and converting the code explanation into semantic vectors; converting a query statement of a user into a query vector, and performing matching search on the query vector and the semantic vector to obtain a matching result; matching results are integrated and fed back to the user interface. The problem that an existing code library retrieval method is low in accuracy and efficiency is solved, a more intelligent code retrieval method is achieved, and the accuracy of large language model questions and answers is improved.

Aspects related to an amalgamation platform providing concealed detection of code-passing using steganography are provided. An amalgamation platform may train a scoring engine to generate suspicion scores for code snippets and an amalgamation engine to generate execution scenarios for the code snippets. The platform may embed code in base code of a network using steganography. The platform may use the steganographic code to detect transmission of a code snippet. The platform may generate a suspicion score for the code snippet. The platform may update the scoring engine based on identifying the suspicion score satisfies a threshold. The platform may generate execution scenarios for the code snippet. The platform may determine a match between an execution scenario and a malicious code scenario. The platform may identify the code snippet as malicious and update the amalgamation engine based on the match. The platform may initiate security actions based on the match.

Embodiments of the present disclosure provide a large language model-based code development method and a related device. Specifically, the capability of invoking a large language model by a code development agent is used to help a user complete a more complex development task in a coding scenario. During conversational interaction between a developer and the large language model, the large language model automatically determines, on the basis of development requirement information and context information provided by the developer, one or more pre-change code files involved in a current round of conversation, automatically generates a code change fragment for each pre-change code file, and, on the basis of the code change fragment corresponding to each pre-change code file, performs code changes on the pre-change code file, so as to obtain a changed code file. For the changed code file generated by the large language model, a manual review mechanism can further be introduced, thereby comprehensively ensuring the quality and reliability of the code. Thus, a more intelligent and collaborative code development solution is provided for developers, and development efficiency and code quality are improved.

The application belongs to the technical field of code automatic generation, and provides a code generation method and system based on a large model. A pre-trained large model is used to convert a natural language description problem into a plan including multiple sequentially executed subtask steps, generate a logical deduction set of the plan and a deduction output set of each test case, compare the logical deduction set and the deduction output set, verify the plan, and correct the plan that fails to pass the verification. According to the problem, the verified plan, and the logical deduction process of each test case, multiple code segments are generated in parallel. The generated code segments are tested using a test case set, the test pass rate of each code segment on the test case set is calculated, the code segment with the highest pass rate is selected as the optimal code segment, and the selected code segment is higher than the set threshold value. The application significantly reduces the required iteration number, and greatly improves the overall efficiency and success rate of code generation.

The application discloses a code base question and answer method and device based on a large model, and relates to the technical field of software engineering. The method comprises the following steps: performing word segmentation on code files to obtain word units, counting the number of word units in each code file, initializing the directory depth of the code files, determining the hierarchical relationship of the code files, generating a code file set, inputting the code file set into a large language model to generate a natural language description set, inputting the code file set and the natural language description set into the large language model to obtain code explanation, segmenting the code explanation and converting the code explanation into semantic vectors, converting a user's query statement into a query vector, matching the query vector with the semantic vectors to obtain a matching result, and feeding back the matching result to a user interface after integration. The method solves the problems of low accuracy and low efficiency of the existing code base retrieval method, realizes a more intelligent code retrieval method, and improves the accuracy of the large language model question and answer.

According to the present disclosure, there is provided a video encoding module comprising an encoder plug-in and a decoder plug-in. The encoder plug-in provides an encapsulator for a first base codec to implement an interface for data exchange with the first base codec to generate an encoded version of an input video signal; the decoder plug-in provides a wrapper for a second base codec to implement an interface for data exchange with the second base codec to generate a decoded version of the encoded video signal generated by the encoder plug-in, wherein the encoder plug-in and the decoder plug-in collectively implement a base coding layer configured to encode and decode a video signal; and the encoder integration layer is used for controlling the operation of the encoder plug-in and the decoder plug-in and controlling the enhancement encoder. The encoder integration layer is configured to: instruct the encoder plug-in to generate the encoded version of the input video signal; instructing the decoder plug-in to generate the decoded version of the encoded video signal generated by the encoder plug-in; and instructing the enhancement encoder to generate an encoded enhancement signal for the encoded video signal, where the enhancement encoder implements an enhancement coding layer, the enhancement encoder configured to: receive the decoded video signal from the decoder plug-in; and generate the encoded enhancement signal for the encoded video signal, the encoded enhancement signal comprising one or more layers of residual data generated based on a comparison of data derived from the decoded video signal and data derived from an input video signal. A method of encoding a video signal, a video encoding system, an encoder integration layer, and an encoder plug-in and a decoder plug-in adapted for use with a video encoding module are also provided.

Embodiments of the present application disclose a code rate estimation method and system based on quantized transform coefficients, a device, and a medium. The technical scheme provided by the embodiments of the present application constructs a weight matrix in advance according to the first coefficient transform result of a plurality of coded transform blocks and the coding code rate. Subsequently, in the case that the coefficient transform and quantization of the current transform block are completed, the weight matrix of the current transform block is selected, and the second coefficient transform result is obtained based on the second coefficient vector of the current transform block and a set index power. Then, the estimated code rate of the current transform block is calculated according to the second coefficient transform result and the weight matrix. By using the above technical means, the code rate calculation efficiency can be improved, and the estimated code rate of the transform block can be quickly determined. Moreover, since the weight matrix is constructed according to the coefficient transform result of the coded module and the coding code rate, the weight matrix has sufficient universality, and the estimated code rate obtained by calculation meets the accuracy requirement of code rate calculation.

This invention discloses a controllable dynamic multi-codebook encryption method and system based on fixed bias weights, belonging to the field of information security technology. Addressing the shortcomings of existing multi-codebook encryption technologies, such as statistical equivalence vulnerabilities, high synchronization overhead, and unquantifiable security strength, this invention constructs a base code pool containing rotated unequal-equivalence codes and uses an independent 64-bit salt-driven Fisher-Yates shuffling algorithm to generate multiple completely independent dynamic codebooks. A non-uniform fixed bias weight is configured for each plaintext symbol to break the statistical correlation between plaintext and ciphertext, and deterministic statistical indistinguishability is achieved by combining bimodal polarization weight allocation. During encryption, codebooks are randomly selected according to weight probability to complete the mapping for each symbol; during decryption, a reverse mapping table is used to achieve fast decryption with O(1) complexity. This invention increases the attack complexity to the product level, achieves a security entropy of 161 bits under standard configuration, has zero communication synchronization overhead, can be deployed on 8-bit MCU devices, and is suitable for low-bandwidth resource-constrained scenarios such as IoT and embedded systems.

This invention discloses a robustness enhancement method for large-scale code models based on context processing and post-training, selecting the enhancement method according to the type of model to be enhanced. For post-training enhancement, adversarial perturbation samples are first generated from the given source code using equivalence semantic transformation, greedy search, and simulated annealing. These perturbation samples are then input into the post-training framework to fine-tune the target large-scale language model. For context enhancement, a backup dataset is generated using backup code. The optimal subset of examples most semantically similar to the current problem is selected by calculating similarity. Dead code is then normalized, and the optimal subset of examples is adaptively selected. Contextual prompts are dynamically adjusted based on task complexity and input into the target large-scale language model. Finally, evaluation metrics and the robustly enhanced target large-scale language model are output. This invention supports multiple mainstream programming languages ​​and both open-source and closed-source models, exhibiting good versatility and scalability.

The invention discloses a code completion model member inference method based on antagonism prompts, which comprises the following steps of: firstly, designing and applying a series of antagonism prompts which keep program functions and are specific to codes, and converting an original input code prefix into a group of disturbed input codes; then querying the target model to respectively obtain prediction outputs of the target model for the original input and the group of input after disturbance, and constructing a feature vector by quantifying the similarity and confusion difference between the two prediction outputs and the real completion code; and finally, training a binary classifier by using the feature vector to automatically deduce the member identity of the sample. According to the method, an inference framework for exposing model memory by utilizing'behavior stability difference 'induced by adversarial prompts is proposed for the first time, an expensive shadow model does not need to be trained, a fine memory mode of a code large language model can be captured, and the accuracy, universality and application feasibility of member inference are remarkably improved.

The invention provides a video coding method and device, electronic equipment, a storage medium, a program product and a method for generating a bit stream. The video coding method comprises the following steps: determining a rate distortion model of each image group in a video sequence; based on the rate distortion model of each image group, adjusting code rate distribution of each image group to obtain a target code rate of each image group; coding the corresponding image group according to the target code rate of each image group; the first step comprises the steps of coding the image group according to a first code rate control parameter value to obtain a first code rate and first distortion; according to the second code rate control parameter value, coding the basic time domain layer in the image group to obtain a basic code rate and basic distortion; obtaining a second code rate and second distortion according to the first code rate, the basic code rate, the first distortion and the basic distortion; and determining a rate distortion model of the image group according to the first code rate, the first distortion, the second code rate and the second distortion. The method can improve the code rate control efficiency.

Aspects related to an amalgamation platform providing concealed detection of code-passing using steganography are provided. An amalgamation platform may train a scoring engine to generate suspicion scores for code snippets and an amalgamation engine to generate execution scenarios for the code snippets. The platform may embed code in base code of a network using steganography. The platform may use the steganographic code to detect transmission of a code snippet. The platform may generate a suspicion score for the code snippet. The platform may update the scoring engine based on identifying the suspicion score satisfies a threshold. The platform may generate execution scenarios for the code snippet. The platform may determine a match between an execution scenario and a malicious code scenario. The platform may identify the code snippet as malicious and update the amalgamation engine based on the match. The platform may initiate security actions based on the match.

This invention discloses a path-oriented diffusion-based code pre-training method and system. The method includes the following steps: 1) Diffusion of existing code training data using a pre-set path diffusion method to obtain a diffused code dataset; wherein, the method for diffusion of a code training data x using the path diffusion method is: occluding or destroying the code training data x to different degrees to obtain n diffusion samples that are related to the code training data x as a diffusion sample set corresponding to the code training data x; the diffusion sample sets corresponding to each code training data constitute the diffused code dataset; 2) Training a code pre-training model using the diffused code dataset. This invention enhances the use of individual code and expands the correlation between multiple training data samples, thereby enabling the trained model to handle downstream tasks more effectively.