97 results about "Similitude" patented technology

The invention belongs to but is not limited to the technical field of artificial intelligence, and particularly relates to a controllable content large model generation method and system based on multi-modal thinking chain reasoning, and the method comprises the steps: S1, improving the capability of multi-agent planning task dependence and execution sequence through employing a thinking chain method, enabling a complex task to be subdivided into a plurality of subtasks through the method, and enabling the subtasks to be subdivided into a plurality of subtasks; a resource transmission process between the sub-tasks is precisely planned and is executed in a chain mode; s2, complex tools are combined into sub-agents with different modal characteristics, so that the number of selectable sub-tasks during task planning is reduced, and the planning capability is improved; and S3, storing scene information of a previous dialogue by means of an external database, and when a new request is encountered, improving correlation and accuracy of multi-agent response by quoting historical interaction information and historical multi-modal resources which have context similarity with the current request, and realizing a more intelligent reaction mechanism.

Methods, systems, and computer program products for dialogue state tracking with in-context tuning are provided herein. A computer-implemented method includes obtaining input dialogue data; identifying, from at least one historical dialogue dataset, one or more historical dialogue examples having at least a given semantic similarity to at least a portion of the input dialogue data; generating, based on the one or more historical dialogue examples, one or more prompts related to at least a portion of the input dialogue data; generating tuning data, associated with at least one dialogue state tracking task related to the input dialogue data, for one or more artificial intelligence techniques by augmenting at least a portion of the prompt(s) in connection with at least one given dialogue state value derived from at least a portion of the historical dialogue dataset(s); and performing one or more automated actions based on the generated tuning data.

Method and system for analyzing at least one computing system for supply chain vulnerabilities of at least one machine learning model include configuring machine learning model and optionally additional data; decomposing operations of the machine learning model's computational graph into smaller decomposed components; associating properties of each decomposed component with properties of the original operations and associating additional data; detecting the semantic similarity of decomposed component and previously encountered decomposed components; converting decomposed components into a standardized representation; calculating signature of decomposed components; evaluating whether portions of the machine learning model are similar to previously calculated signature; testing for supply chain and model vulnerabilities that exist based on previous signatures; identifying vulnerabilities that persist and correlating defenses; storing the generated signatures, identified vulnerabilities, and identified defenses; and generating report detailing the machine learning model's supply chain, vulnerabilities, and defenses.

Data heterogeneity has been shown to reduce the convergence speed and model accuracy in existing federated learning (FL) and distributed learning (DL) based artificial intelligence or machine learning (AI / ML) model training processes. In some embodiments, a group of devices participating in an FL or DL based AI / ML model training process is divided into clusters, e.g., based on similarities in their data class distributions, and each device is configured with a cascaded AI / ML model that includes a first AI / ML sub-model and a second AI / ML sub-model that are cascaded. For each device, one of the cascaded sub-models is a common AI / ML sub-model that is common to all devices in the group, and the other sub-model is a cluster-specific AI / ML sub-model that is common to the cluster that the device has been assigned to within the group of devices. A multi-stage training process for the cascaded AI / ML model is also provided.

A system is configured to: (a) provide a deep learning model; (b) receive a concept, the concept including text; (c) determine a word embedding representing the concept using the deep learning model; (d) determine a centroid for the concept, the centroid representing an average position of concept characteristics in a high-dimensional word embeddings space; (e) determine Euclidean distances for each of the concept characteristics; (f) determine cosine similarities for each of the concept characteristics; (g) determine a prominence index for each of the concept characteristics based on the determined Euclidean distances and the determined cosine similarities; and (h) provide, to a client device, a ranking of the concept characteristics based on the prominence index.

The invention provides a self-adaptive template reconstruction method and system based on a small amount of samples, and belongs to the field of natural language processing, and the method comprises the steps: S1, constructing an initial prompt template based on a small amount of sample data; s2, inputting the initial prompt template into a large language model to generate a question pool; s3, generating an answer pool by using a large language model; s4, performing screening and clustering analysis on the answers in the answer pool through a multi-level similarity comparison module, and dynamically adjusting weight parameters of similarity calculation in the multi-level similarity comparison module through a model self-feedback mechanism to obtain an overall score of the screened answers; s5, calculating a comprehensive similarity score of each answer group, performing expert scoring on each answer group, and reconstructing the initial prompt template based on the comprehensive similarity score and the expert score; and S6, inputting the reconstructed prompt template into a large language model for iterative updating until a final answer is obtained.

The invention discloses a test question management method based on multi-modal adaptive similarity learning, and relates to the technical field of data management, and the method comprises the steps: collecting an original test question set, generating an evolutionary test question set through a preset large language model, generating a variable test question set through a data enhancement technology, and carrying out the fusion processing of the test question set to form a test question ecological library, the diversity and coverage of test question resources are enriched; the test question ecological library is input into a preset semantic similarity model and a preset syntactic similarity model to obtain a semantic similarity score and a syntactic similarity score of each test question, so that the similarity of the test questions can be comprehensively measured, and the screening accuracy is improved; the multi-modal similarity score is obtained according to the semantic similarity score and the syntactic similarity score to be compared with the preset threshold value, then the test questions are imported into the auditing module or the question bank storage module according to the comparison result, intelligent distribution of the test questions is facilitated, and the timeliness and quality of the question bank are ensured.

Adaptive verifiable training enables the creation of machine learning models robust with respect to multiple robustness criteria. In general, such training exploits inherent inter-class similarities within input data and enforces multiple robustness criteria based on this information. In particular, the approach exploits pairwise class similarity and improves the performance of a robust model by relaxing robustness constraints for similar classes and increasing robustness constraints for dissimilar classes. Between similar classes, looser robustness criteria (i.e., smaller ∈) are enforced so as to minimize possible overlap when estimating the robustness region during verification. Between dissimilar classes, stricter robustness regions (i.e., larger ∈) are enforced. If pairwise class relationships are not available initially, preferably they are generated by receiving a pre-trained classifier and then applying a clustering algorithm (e.g., agglomerative clustering) to generate them. Once pre-defined or computed pairwise relationships are available, several grouping methods are provided to create classifiers for multiple robustness criteria.

The invention belongs to the technical field of artificial intelligence and natural language processing, and discloses an NL2SQL model training and storing method and device based on a GRPO reward function. According to the method, a high-quality training data set is constructed, and the training effect and the generation performance of the model are effectively improved in combination with grammar verification, execution verification and semantic consistency screening; according to the method, the GRPO is adopted as a basic framework, and a combined reward function with multiple dimensions of coverage execution accuracy, grammar legality, semantic similarity, mode link and the like is designed; according to the method, a staged course learning strategy is designed, the sub-reward functions are activated and adjusted in a staged mode, the model is guided to be gradually transited to semantic understanding and execution optimization from structure specifications, and the generalization ability is improved; meanwhile, a reward weight dynamic adjustment mechanism is introduced, the weight of a reward function is automatically adjusted when SQL query execution fails, and the stability and sensitivity of training feedback are enhanced.

The invention discloses a multi-feature fusion binary code similarity detection method based on a large language model, and belongs to the technical field of information security. Compared with a conventional detection method which only depends on a single feature, the method has the advantages that fine grammar and global structure information of a function can be bidirectionally captured by utilizing a large language model which is finely adjusted by prompt word engineering and a GraphCNN graph convolutional neural network, so that the comprehensiveness and accuracy of feature expression are greatly improved; meanwhile, a hierarchical Navigation Small World Graphs (HNSW) algorithm is applied to efficiently retrieve candidate similar functions in an offline vector library, so that the method is suitable for large-scale database application; and finally, through fine comparison of the Siamese network, similarity detection with higher accuracy is realized under the condition of low cost. According to the method, through multi-dimensional fusion of semantics and structure information, the accuracy and efficiency of similarity detection are remarkably improved, and a more accurate and robust solution on the basis of low cost is provided for binary code similarity detection.

The invention discloses a multi-dimensional example selection method and system applied to a numerical reasoning task, and relates to the technical field of natural language processing, and the method comprises the steps: employing a plurality of demonstration problems and corresponding demonstration code thinking chains to form a plurality of training demonstration examples; dividing an ordered positive example group and an ordered negative example group by adopting a large language model based on the prompt answer accuracy determined by each training demonstration example; performing model training based on the training query problem, the ordered positive example group and the ordered negative example group through a to-be-trained example scoring model according to similarity, uncertainty and complexity, and determining a trained example scoring model; when the target query problem is determined, candidate demonstration scores of candidate demonstration examples corresponding to the target query problem are determined through the trained example scoring model, and a target demonstration example is selected. On the basis of the scheme, the programmed thinking is introduced to generate examples to perform multi-dimensional collaborative evaluation on the examples, and the example scoring model is optimized through inter-group differences, so that the reliability of example selection is improved.

The invention discloses a similarity comparison-based large model supply chain automatic restoration method and device, and the method comprises the steps: firstly carrying out the semantic coding of an original vulnerability code through a pre-training model, and constructing a semantic and structure parallel dual-channel representation in combination with an abstract syntax tree and other program structures; cWE type intelligent classification is performed on vulnerabilities by using a locally deployed large language model subjected to LoRA fine tuning, and meanwhile, a zero sample semantic matching mechanism is introduced, so that the recognition capability of unknown vulnerability types is improved. And searching the closest historical case from the knowledge base through similarity vector comparison, and extracting a repair abstract to construct a model to generate a prompt. Patch codes and repair instructions are generated through a large model, automatic filing and knowledge base updating are supported, and the continuously-enhanced automatic repair capacity is achieved. The method can be widely applied to automatic vulnerability repair scenes of supply chain components such as large model plug-ins, code interfaces and dependent packages, and the safety, functionality and interpretability of code repair patches are greatly improved.

Embodiments classify a product to one of a plurality of product classifications. Embodiments receive a description of the product and create a first prompt for a trained large language model (“LLM”), the first prompt including the description of the product and contextual information of the product. In response to the first prompt, embodiments use the trained LLM to generate a hallucinated product classification for the product. Embodiments word embed the hallucinated product classification and the plurality of product classifications and similarity match the embedded hallucinated product classification with one of the embedded plurality of product classifications. The matched one of the embedded plurality of product classifications is determined to be a predicted classification of the product.

The invention discloses a code review comment generation method and system based on reinforcement learning, belongs to the technical field of crossing of software engineering and artificial intelligence, and aims to solve the technical problem of how to improve the accuracy, efficiency and expandability of code review and overcome the defects that an existing tool is low in comment quality and poor in practicability. According to the technical scheme, the method comprises the steps of collecting and preprocessing data of code differences, human comments and real correction codes, and constructing a data set; code review comments are generated, specifically, a pre-trained large language model LLM is finely adjusted through the data set, code difference data are collected, and the review comments are obtained; semantic similarity reward: calculating the semantic similarity between the generated comment and the real comment, and generating a reward signal Rsematic; the generated comment and code difference is input into a code optimization model, a corrected code patch is generated, the similarity between the generated patch and a real patch is evaluated, and a reward signal Rtask is generated; performing reinforcement learning and fine tuning on the large language model LLM; and deploying and generating comments.

The invention discloses a multi-modal large model assisted noise robust human skeleton unsupervised representation learning method and device, and the method comprises the steps: carrying out the sampling of an RGB video corresponding to a human skeleton sequence, obtaining a key frame, and generating an action text description through a multi-modal large model; the human body skeleton sequence and the corresponding action text description are encoded to a public space, a more reliable learning target is generated through a sample noise evaluation method of abnormal sample and information entropy joint estimation, and a model is trained by using a noise perception contrast learning method; and finally, an encoder with better performance can be obtained after a certain batch of iterative training. According to the method, on the premise that no extra training reasoning cost exists, text description irrelevant to actions is distinguished through the capacity of the encoder, so that the model learns more reasonable similarity representation, and meanwhile the performance of the encoder in downstream tasks is well improved.

There are provided systems and methods of an artificial intelligence system for identifying, evaluating, and controlling overlapping rule code. A service provider, such as an electronic transaction processor for digital transactions, may utilize computing services that implement coded rules and rule-based engines for decision-making of data including real-time data processing in production computing environments. Multiple rules may overlap, and the service provider may utilize an AI system to identify the overlap by converting rule code and logic to syntax in a language. The syntax may then be analyzed for overlap and a similarity score calculated for pairwise similarity between each rule in the rule-based system. Thereafter, based on similarity scores and pairwise similarity, clusters of rule pairs may be identified in order to reduce rules through merging or deleting the same, similar, or overlapping rules based on their syntaxes.

The invention relates to the technical field of intelligent education systems, in particular to a subjective question automatic correction method based on multi-technology fusion, which comprises the following steps of: firstly, identifying whether a question is a strong proposition or not through a regular expression, and then performing knowledge point extraction, off-question judgment and double similarity evaluation on the strong proposition question; plagiarism detection, cosine similarity calculation and user-defined rounding are carried out on all the questions, and finally personalized comments are generated and correction results are output. According to the method, through double similarity evaluation, semantic relevance between answers and standard answers as well as questions is comprehensively considered, the misjudgment rate is remarkably reduced, meanwhile, through strong proposition recognition, knowledge point extraction and double similarity evaluation, the pertinence and accuracy of correction are remarkably improved, whether the answers of students meet the requirements of the questions or not can be deeply analyzed, and the correctness of the questions is improved. Misjudgment caused by insufficient semantic understanding in the prior art is avoided, and the method is particularly suitable for accurate scoring of strong proposition questions in subjective questions.

The invention relates to a function point duplicate checking method, a function point duplicate checking device and function point duplicate checking equipment based on large language model application, which are used for judging whether two function points in system or software development are duplicated or not by taking coincidence and similarity of the two function points on an operation dimension and an object attribute dimension as standards. And during the period, a current obtained result is evaluated through a reflection mechanism, and a more complete duplicate checking result is obtained through deep analysis. And whether the two function points are repeated or not is judged by combining the two dimensions of operation and object attributes, so that the problem that whether the two function points are repeated or not cannot be effectively judged in the traditional technology is solved. The Reflexion framework is referred to, and the ReAct mechanism-based agent is created, so that self-reflection can be performed while judgment is generated, self output is analyzed and evaluated in real time, and the judgment accuracy and adaptability are effectively enhanced.

A system configuration derivation device calculates a first evaluation value indicating an evaluation of a similarity between a specific embodiment rule application and a candidate of the embodiment rule application, wherein the specific embodiment rule application is included in a design history, the design history indicates a history of in a case where an embodiment configuration is obtained by repeatedly applying an embodiment rule to an abstract configuration; selects, based on a second evaluation value, a specific abstract configuration among the abstract configurations obtained by applying the embodiment rule to the new configuration one or more times, wherein the second evaluation value indicates the evaluation of the similarity with the design history, of the entirety of one or more applications of the embodiment rule until the abstract configuration is obtained from the new configuration; and repeats the embodiment rule application to the selected specific abstract configuration.

The invention belongs to the technical field of federated learning security defense, and particularly relates to a backdoor attack defense method based on model parameter similarity detection, a program, equipment and a storage medium. According to the method, the change of the updating amplitude and the output of each model is analyzed to reflect whether client data sample distribution changes or not, so that abnormal model updating is found. According to the method, records of historical model updating amplitude and historical output classification results are maintained for each model, the average similarity between the historical model updating amplitude and the historical output classification results is calculated, whether data sample distribution of a client changes or not is comprehensively evaluated based on the average similarity, and a model parameter normalization technology is combined to determine whether the data sample distribution of the client changes or not. Back door poisoning updates may be filtered while preserving benign model updates.

Methods and systems for generating output content using a generative artificial intelligence (AI) model based on an input. A similarity-assessment layer at the output of the generative AI model determines a similarity measure for the output content vis-à-vis pre-existing items in a repository. The similarity measure is compared to a threshold value and, responsive to the comparison indicating excessive similarity, one or both of the input and the generative AI model are adjusted, and the generative AI model is re-run to generate new output content.

The invention discloses a method and system for evaluating and improving paragraph similarity judgment precision of a large model, and belongs to the technical field of artificial intelligence and natural language process.The method comprises the steps that a scoring standard and an evaluation process are formulated, the scores account for 50% of the overall score; scoring by experts in the three fields of business, algorithm and test according to indexes; automatically scoring the paragraphs of the test set by adopting the trained large model, comparing with manual comprehensive scoring, considering that the model is accurate when the difference does not exceed 0.5, and calculating the accuracy rate of the model according to the result; model optimization and promotion: constructing a training set; carrying out model training and iteration; preparing a test set and accurately evaluating the test set; and performing cyclic feedback optimization. According to the method, the accuracy and practicability of the large model in paragraph similarity judgment can be improved, and the method is more reliable and efficient in practical application.

Methods and systems for generating output content using a generative artificial intelligence (AI) model based on an input. A similarity-assessment layer at the output of the generative AI model determines a similarity measure for the output content vis-à-vis pre-existing items in a repository. The similarity measure is compared to a threshold value and, responsive to the comparison indicating excessive similarity, one or both of the input and the generative AI model are adjusted, and the generative AI model is re-run to generate new output content.

THE WAY IN WHICH THE ESSENTIAL ASPECT OF THE STRATEGE "FROM NOTHING - SOMETHING" IS IMPLEMENTED IN INFORMATION TECHNOLOGIES. The initial essence "Nothing" (0, Zero) of the method is the abstract essence. The mathematical model of the Projected Object Entity (PEO) is embodied in material, software, and mixed entity objects from various disciplines—the "something" of the method. The methodology of comparative cause-and-effect analysis is applied in the creation of the PEO, and the methodology of essential synthesis is used in its design: subject-specific and essential interdisciplinary properties of similarity, universal mathematical models in the form of orgraphs and their matrices, graphs with algorithms of nodes and their matrices, as well as essential identities, equations, and differential equations that reflect the balance of actions and counteractions.And since the professional and interdisciplinary properties of similarity for individual objects of different disciplines continue to apply, the transfer of the parental properties, qualities and the collective image of "nothing" and its "something" takes place "according to the image and likeness" of the essence.

The invention provides an active man-machine cooperation method driven by process knowledge and interactive semantics, and the method comprises the steps: carrying out the step-by-step optimization and updating of a general task description text and a dynamic assembly task interaction description text in the manufacturing field, and obtaining an initial generation result; performing evaluation according to a semantic similarity classification standard and a real label to obtain a final generation result, performing observation sampling on a posture sequence when an operator implements an assembly behavior according to the generation result to obtain observation postures and dynamic scene semantic information, and extracting features to obtain a diffusion matrix; further processing to obtain an inherent matrix and a weight matrix to form point matrix features, mapping the point matrix features to a Gaussian distribution coding set feature map, sampling potential coding factors, mapping the sampling potential coding factors to a posture sequence to obtain a current posture sequence, predicting to obtain an operator assembly behavior prediction result, and combining an assembly environment state and an assembly task state to obtain an operator assembly behavior prediction result. Intelligent agent actions and operator assembly behaviors are evaluated, current state and reward function feedback are provided for the intelligent agents, and the robot is controlled to move.

The invention discloses a fine-grained class incremental learning method and system based on a decoupling Gaussian classifier, a storage medium and equipment, and belongs to the field of reinforcement learning of artificial intelligence. The method comprises the following steps: decoupling different types of features extracted by a backbone network into a plurality of independent Gaussian distributions, calculating the similarity among different Gaussian distributions through regularization penalty, encouraging to minimize the overlapping among similar features, and increasing the inter-class distance. According to the method, forgetting in fine-grained incremental learning is analyzed from the perspective of examples, and a universal Gaussian decoupling classifier is provided, so that the distinguishability of similar classes in the incremental process is enhanced, and the classification precision is improved.

The invention provides a case reasoning-based mechanical equipment intelligent design method. The method comprises the following steps: constructing a mechanical equipment design case library; calculating subjective and objective weights of design features, optimizing subject and guest weight distribution, and calculating a mixed weight; taking the new problem as a retrieval condition, and identifying candidate cases based on weighted similarity; normalizing problem attributes of the candidate cases according to column maxima to form a similarity matrix; solving utility values of candidate case problem attributes to form utility vectors; normalizing problem parts of the candidate cases according to column maxima to form a similarity matrix; combining the similarity matrix and the grey incidence relation matrix, and obtaining a weighted average matrix after normalization; and solving the value of the solution feature of the new problem by using a weighted average method according to the elements of the weighted average matrix so as to obtain a solution. According to the invention, the efficiency and the intelligent level of mechanical equipment design can be improved.

The application discloses a small sample face recognition method based on prototype calibration and adaptive fine-tuning, comprising the following steps: 1) pre-training the encoder on a large-scale face database, capturing rich semantic information and class distribution, constructing an evaluation set, and ensuring that the training identity information does not overlap with the constructed evaluation set; 2) initializing the classifier weight through the weight imprint method, ensuring smooth transition, and inducing the discriminative structure of the encoder; 3) calculating the cosine similarity between the fine-tuning class prototype and the pre-training prototype, fusing the pre-training class information through weighted average, and calibrating the biased fine-tuning class prototype; 4) introducing an adaptive margin loss function, adjusting the sample penalty strength, and selectively fine-tuning the BatchNorm layer and the classifier weight; and 5) using the field-aware similarity NAC to calibrate the unknown class rejection, enhancing the similarity relationship with the field sample, and obtaining the final matching score. The method can improve the recognition accuracy and rejection ability of the model in the small sample learning scene.

The invention provides a code similarity review method based on artificial intelligence, and belongs to the field of artificial intelligence and software engineering.The code similarity review method comprises the steps that semantic features of codes are extracted through a deep learning model, code similarity classification is conducted through a machine learning model, the semantic features of the codes can be automatically learned, and the similarity of the codes is detected; the problems of potential code plagiarism, code reuse, code vulnerabilities and the like are identified, and the accuracy, efficiency and expandability of code similarity review are improved.

Comprising a method and apparatus including computer code configured to cause one or more processors to construct a similarity score between a plurality of words, initialize a similarity structure that depends on a task and is based on the similarity score, and initialize the similarity structure based on the similarity score by implementing a bilayer optimization. And performing machine learning on the task dependency of the similarity structure, the double-layer optimization comprising: a search stage comprising learning a model weight by estimating a model parameter corresponding to a first entry of the similarity structure, learning a parameter of a second entry of the similarity structure by applying the model parameter to the second entry; and a fine tuning stage: updating the model parameters under the condition of keeping the similarity structure unchanged.