1914 results about "Large model" patented technology

A text sequence recommendation method and system based on a large language model is disclosed, belonging to the technical field of recommendation algorithms. The method includes: a data preprocessing stage, a large language model pre-training stage, a sequence model fine-tuning stage and a matching stage. According to this disclosure, a large language model is introduced into a text sequence recommendation task, so that text can be better modeled by utilizing rich pre-training corpus of the large language model; meanwhile, sequence modeling is performed on the text, the capability of sequence recommendations modeling in a large model is activated, an ID-based recommendation paradigm in a traditional recommendation algorithm is eliminated, and recommendation task learning processing is better performed in a cold start scenario and a knowledge transfer scenario; and finally, a recommendation result is finally optimized by a sequence model.

The application discloses a government affair text auxiliary writing system and method based on a multi-modal large model, and relates to the technical field of artificial intelligence and informationization; comprising: step 1: converting heterogeneous government affair data into unified semantic representation, realizing cross-modal deep alignment through fine-grained contrast learning; encoding text, image and voice features by using an encoder respectively, and constructing a training set containing 100,000 pairs of government affair image-text samples; weighting and fusing the text, image and voice features, dynamically adjusting the weight through an attention mechanism, training a large model, step 2: performing lightweight deployment of the large model, and step 3: using the large model to understand and identify user intention, and outputting government affair text.

The invention discloses a strategic recommendation method and device, and relates to the technical field of intelligence analysis tools, the method is applied to a computer device capable of calling a knowledge base and a large model, and the method comprises the following steps: retrieving an obtained text fragment based on an input request of a user; based on the input request, calling a large model to carry out site analysis on the input request to obtain a site analysis result; based on the input request, the text fragment and the standing analysis result, calling the large model to combine with the text fragment to perform qualitative analysis on the standing analysis result, and outputting a qualitative analysis result; based on the input request and the qualitative analysis result of the dictionary format, determining an optimal feature vector of the dictionary format by using an analytic hierarchy process; and on the basis of the optimal feature vector, four-quadrant analysis data of each strategy is calculated by using a four-quadrant analysis method, and then optimal strategy pushing is completed. According to the method, qualitative analysis, an analytic hierarchy process and a four-quadrant analysis method are combined, development strategic intelligence analysis is carried out according to requirements input by a user, and strategic suggestions are generated.

This invention provides a virtual reality method and system for constructing a geometrically perceptible material field based on a large model and driving real-time interactive sound synthesis. The method includes: acquiring geometric data and multi-view observation data of the target object; generating prior material constraint information from a large model and fusing the multi-view data to obtain the material feature field of the target object; generating physical sound assets offline based on the material feature field; acquiring interaction state information during virtual reality scene runtime, mapping it to excitations consistent with geometric position and interaction process, and synthesizing collision and friction sounds, enabling the same object to produce continuously changing and distinguishable sound feedback under different touch positions, sliding paths, and contact areas. This invention achieves real-time synthesis and output of collision and friction sounds by constructing a material feature field bound to the geometric space of the virtual object and driving physical sound assets in conjunction with interaction state information during runtime, thus enhancing the consistency of auditory feedback and immersive experience in virtual reality scenes.

The application provides a cloud-based industrial equipment multi-task integrated operation and maintenance method and system. The method comprises the following steps: obtaining an operation and maintenance task of an industrial equipment, constructing a multi-task integrated operation and maintenance basic model based on the operation and maintenance task; obtaining multi-modal feature data of the industrial equipment, and presetting a cross-modal feature fusion mechanism based on the multi-modal feature data; training the multi-task integrated operation and maintenance basic model by using a multi-modal cloud joint method based on the preset cross-modal feature fusion mechanism; obtaining a multi-task integrated operation and maintenance large model of the industrial equipment based on the trained multi-task integrated operation and maintenance basic model; and performing multi-task integrated operation and maintenance of the industrial equipment, such as anomaly detection, fault diagnosis and life prediction, based on the multi-task integrated operation and maintenance large model. The multi-task integrated operation and maintenance large model formed by the application significantly improves the operation and maintenance efficiency and reliability of the industrial equipment in a complex scene.

This invention relates to a multimodal visual understanding method based on consistency learning and hybrid feature extraction. It includes constructing an end-to-end fine-grained consistency learning framework, introducing a hybrid region extractor, fusing local details and global semantics to generate high-quality hybrid visual cue embeddings, combining self-reconstruction loss and latent spatial consistency loss to force the model to establish explicit alignment between the input visual cue and the output segmentation label, utilizing the geometric boundary constraints of the localization task for description generation, and simultaneously optimizing localization accuracy using the semantic depth of the description task. Furthermore, it constructs a detailed localization index expression and segmentation task to enhance the model's reasoning ability for complex long text instructions. The aim is to address the problems of feature fragmentation and insufficient accuracy in existing large models for fine-grained visual localization and description tasks. Compared with existing technologies, this invention has advantages such as high accuracy and strong generalization ability in pixel-level localization and fine-grained description.

The disclosure provides a large model-based quality inspection method, device, equipment, storage medium and program, relating to the technical field of data processing, specifically relating to artificial intelligence, deep learning and large language model technology, which can be applied in the field of AI medical technology, wherein the method comprises: obtaining to-be-inspected data of a target platform and a target quality inspection prompt template; wherein the target quality inspection prompt template comprises quality inspection task instructions and quality inspection question classification content; splicing the to-be-inspected data and the target quality inspection prompt template to obtain a target quality inspection prompt; inputting the target quality inspection prompt into a target large model to automatically inspect the to-be-inspected data according to the target quality inspection prompt through the target large model, and obtaining first quality inspection data. The embodiment of the disclosure can reduce the artificial quality inspection cost of platform quality inspection and improve the efficiency and accuracy of platform quality inspection.

The application provides a visual language large model adaptive bias evaluation method for evaluating bias performance of a target model on a specific task, and the method comprises the following steps: S1, obtaining an initial evaluation dataset, wherein the initial evaluation dataset comprises a plurality of evaluation data; S2, fine-tuning a preset large language model using the initial evaluation dataset in a preset fine-tuning manner to obtain an initial data generator; and S3, generating evaluation data based on the initial data generator to perform multiple rounds of bias evaluation on the target model, wherein in each round of bias evaluation, based on all evaluation data corresponding to the last round of bias evaluation, a data generator corresponding to the last round of bias evaluation is iteratively updated according to a preset objective function to obtain a current round data generator, and the current round data generator is used to generate a current round evaluation dataset to perform bias evaluation on the target model.

The application discloses a domain knowledge triple extraction method, system, medium and equipment, and relates to the technical field of knowledge engineering.The domain knowledge triple extraction method and system provided by the application can provide high-quality and efficient data support for triple extraction by analyzing and converting the obtained domain professional text into a structured vector knowledge base which can be efficiently searched, then the domain professional text is subjected to quantitative calculation and fusion screening, the initial entities of domain core high-frequency words which have global high frequency and cross-text universality are mined, the problem of "local optimum and global deviation" caused by traditional single word frequency is avoided, non-core term interference is effectively avoided, and the stability, relevance and efficiency of iterative extraction are improved, and further, the high-frequency word initial entities are subjected to closed-loop iteration through iterative RAG algorithm combined with the structured vector knowledge base, accurate extraction of the large model and iterative correlation entities, the explicit and implicit semantic correlation triples in the text can be mined layer by layer, the strong relevance and structural integrity of the extracted triples are ensured, and the professionalism and accuracy of the model in extracting triples are improved.

This invention proposes a method and system for judging tunnel defects based on large-scale model identification of tunnel point cloud data, belonging to the field of tunnel engineering inspection. The method includes: acquiring the original 3D point cloud data of the tunnel and preprocessing it into a standardized point cloud dataset; segmenting the standardized point cloud dataset and constructing image multimodal data pairs; inputting the image multimodal data pairs and dedicated prompting instructions into the constructed defect detection large-scale model to obtain preliminary identification and evaluation results; mapping the results back to the original 3D point cloud coordinate system through a view projection matrix to construct a 3D defect point cloud; constructing a temporal-spatial knowledge graph based on the 3D defect point cloud, generating early warning information when preset risk conditions are met; and generating tunnel defect detection results based on the 3D defect point cloud and early warning information. This invention achieves full-process automation and intelligence in tunnel defect detection, significantly improving detection efficiency and accuracy.

This invention relates to the field of artificial intelligence education technology, specifically to a method and system for reviewing and providing feedback on physics, chemistry, and biology lab reports based on a multimodal large-scale model. The method includes: acquiring lab report question data, student answer data, and scoring standard data; vectorizing the question data and retrieving scoring criteria from a scoring knowledge base; selecting either a text-based or image-based scoring process based on the answer type, and calling a fine-tuned multimodal large-scale language model for scoring reasoning; calling an error identification module and a teaching suggestion generation module based on the scoring results; summarizing the complete review results and calculating the joint confidence level; and determining whether to push the results for manual review based on the confidence level. This invention can improve the accuracy, fine-grained understanding, and quality of teaching feedback in lab report review, enabling the self-evolution of the review system.

The application provides a data processing method and system based on a multi-modal large model, relates to the technical field of data processing, and comprises the following steps: analyzing and identifying to-be-inspected data, and determining corresponding engineering types and inspection strategies; sending a query request to a pre-trained knowledge graph and a pre-trained vector knowledge base based on the engineering types and the inspection strategies, obtaining path retrieval results associated with structured knowledge and semantic retrieval results associated with unstructured knowledge, and fusing the path retrieval results and the semantic retrieval results to obtain fused retrieval results; the pre-trained knowledge graph comprises entities, attributes and the association relationships between the entities; the pre-trained vector knowledge base comprises a plurality of professional material text coding blocks; the to-be-inspected data is input into a pre-set inspection model for preprocessing and fine-grained element recognition, global macroscopic description, key path calculation and key information extraction are completed, reasoning is performed in combination with the fused retrieval results, and a structured inspection result is generated.

The application discloses a reliability evaluation and optimization method and system for large model function code, a medium and a program product, and belongs to the technical field of artificial intelligence. A demand prompt word, a test case and expected output are acquired to form a standardized test data set; an execution consistency score, a semantic consistency index and a static structure consistency index of the code to be evaluated are calculated and fused into an intention consistency comprehensive score; a safety score is obtained by performing static safety analysis on the code to be evaluated; the intention consistency comprehensive score and the safety score are fused into a final comprehensive score, and the reliability is judged; if the reliability is not good, a defect label is generated according to the scores of the indexes and weighted fusion is performed to obtain optimization guidance information, and the code is corrected or regenerated according to the information. The application comprehensively evaluates the code quality from four dimensions of execution, semantics, structure and safety, highlights the short board effect by adopting a fusion strategy, realizes closed-loop automatic optimization, and significantly improves the reliability and safety of the function code.

The present application relates to a method and apparatus for training a mixture-of-experts model, a computer device, a computer-readable storage medium, and a computer program product. The method comprises: acquiring a training sample set of a mixture-of-experts model associated with input to a large model, the training sample set comprising input data of the large model and input parsing label information corresponding to the input data; inputting a target sample in the training sample set into the mixture-of-experts model, and obtaining activation probability information of each expert model in the mixture-of-experts model, a target output result of the mixture-of-experts model, and a sub-output result of a loaded expert model; obtaining expert differentiation reward information on the basis of model parameters of each expert model, the target sample, the target output result, and the sub-output result; and performing iterative training on the mixture-of-experts model on the basis of the activation probability information, the expert differentiation reward information, the target sample, and the target output result, and obtaining a trained mixture-of-experts model.

The application provides a substation civil engineering large model training method and system integrating design specifications, and belongs to the field of substation civil engineering design. The method comprises the following steps: constructing a unified substation structure diagram; obtaining a substation civil engineering design specification text and parsing it into design rule embeddings, including trigger functions, constraint functions, rule weights, and repair templates; inputting the structure diagram into a multi-modal large model to generate an initial design output representation; performing a differentiable constraint projection process on the initial design output representation based on the trigger functions, constraint functions, and rule weights to obtain a projected design output representation; verifying the compliance of the projected design output representation with the design specifications based on a symbolic verifier, and generating corresponding rule violation counterexample information when the verification fails; performing a repair generation operation on the rule violation counterexample information based on a repair generator to obtain a repair candidate design, and backflowing the repair candidate design for large model training. The application can enable the design specifications to play a constraint role in the model generation and training stages.

The application belongs to the technical field of large model security, and specifically discloses a large model threat sample library construction method, which comprises the following steps: step S1, embedding malicious instructions or malicious data into prompt words to construct threat samples; step S2, inputting the threat samples into a target large model as prompt words; step S3, using an evaluation large model to evaluate the output results of the target large model; and step S4, if the evaluation result is harmful content, adding the threat sample into a threat sample library, otherwise, discarding the threat sample. The application solves the problem of the lack of a threat sample library for security testing of an existing large model, and automatically tests and evaluates the threat samples during the construction of the sample library, so that the threat samples can be stored in the library only after the evaluation, thereby effectively improving the construction efficiency of the threat sample library. Meanwhile, the constructed threat sample library can support the security testing of the target large model, and can be used to improve the security of the target large model.

This invention discloses a method for autonomously constructing an evolvable knowledge graph by integrating a multimodal large-scale model, relating to the fields of artificial intelligence and knowledge graph technology. The method includes: collecting raw multimodal data from text, images, audio, and structured tables; using a pre-trained multimodal understanding large-scale model to perform unified semantic encoding on the raw multimodal data, generating modality-independent deep feature vectors to form an initial multimodal feature pool; performing cross-modal clustering analysis on the vectors in the feature pool, grouping semantically similar vectors into the same feature cluster, with each feature cluster defined as a candidate knowledge concept node; constructing an initial concept relationship network based on the co-occurrence relationship and feature similarity between candidate concept nodes; iteratively optimizing and evolving the network according to preset graph quality evaluation indicators, outputting the final evolvable knowledge graph. This invention achieves the autonomous and unified construction and evolution of a knowledge graph from multimodal data.

This invention discloses a method for mining alumni information based on a large-scale model, comprising the following steps: 1) acquiring raw alumni-related data; 2) preprocessing the acquired data; 3) performing natural language processing on the acquired data; 4) semantic parsing and information extraction using a large-scale model; 5) constructing a knowledge graph; storing the extracted alumni information and relationships in a graph database; and constructing a visualized alumni knowledge graph; 6) updating the knowledge graph data; and 7) providing intelligent retrieval and question-answering services for alumni information based on the knowledge graph data. This invention uses a large-scale model to perform deep semantic understanding and information extraction from unstructured text. Compared to traditional keyword-based or small-scale model-based methods, it can more accurately identify alumni identities, professional experiences, and social relationships, while reducing the false matching rate.

The application discloses a kind of multi-dimension resume intelligent screening and post matching method based on large model, it is related to artificial intelligence and recruitment technical field.The method includes: resume and post information are preprocessed and standardized;Using recruitment field fine-tuning large model, combined with knowledge graph, hierarchical attention mechanism and BiLSTM-CRF structure, extract the multi-dimension feature of resume and post;By cosine similarity, analytic hierarchy process and hungarian algorithm, construct matching model, realize global optimal matching;Adaptive threshold is used to complete final screening and output matching result.The application can significantly improve resume screening efficiency and matching precision, solve the problems of traditional mode such as insufficient semantic understanding, one-sided feature extraction and simple matching logic, suitable for various enterprise recruitment and human resource management scenarios, with high practicality and popularization value.

This application discloses an AI multi-turn question answering evaluation method, device, and storage medium. It relates to the field of large model corpus evaluation technology. The method includes: constructing a global contextual feature pool based on the global task obtained from parsing input information and the baseline evaluation value of the domain to which the global task belongs; responding to the output text of the large language model in the nth round, evaluating the output text based on a pre-set first model and a second model to obtain an absolute quality evaluation value; comparing the output text with the global contextual feature pool corresponding to the (n-1)th round based on a pre-set dimension to obtain a relative incremental evaluation value; and weightedly fusing the absolute quality evaluation value and the relative incremental evaluation value according to the completion degree of the global task in which the output text is located to obtain a text evaluation value. This application can accurately identify homogeneous and worthless corpora in AI multi-turn generation, improving the quality identification accuracy and value screening efficiency of AI multi-turn question answering generated content.

This invention discloses an AI workflow intelligent orchestration method and system based on a large model, comprising: receiving a user-input task description, input data, and target constraint information; performing semantic parsing on the task description using a large model to obtain a task intent model; decomposing the overall task into multiple atomic task nodes according to the task intent model, and generating a candidate workflow topology including atomic task nodes and logic control nodes; matching target tool components for each atomic task node in a pre-built tool library and initializing node parameters based on the functional requirements of each atomic task node; performing a pre-evaluation of the candidate workflow topology, and when the evaluation result is lower than a preset threshold, correcting the nodes, connections, parameters, or control logic to obtain the target workflow; and executing the target workflow to obtain the target result. This improves the orchestration flexibility, result quality, and execution reliability in complex task scenarios.

This application discloses an intelligent retrieval control method, device, and electronic device based on large model dimensionality reduction, relating to the field of artificial intelligence technology. The method includes: inputting the text to be retrieved into a large language model as an embedding model, and obtaining the output vector of a specified hidden layer of the large language model to obtain a text vector to be retrieved, the dimension of which is M; using a specified window to perform dimensionality reduction processing on the text vector to be retrieved, reducing the text vector to N-dimensionality reduced text vectors, the window size of which is determined based on M and N; dividing the text vector to be retrieved into N sub-vectors by the specified window; in the text vector to be retrieved, the corresponding element positions of adjacent sub-vectors in the N sub-vectors are at least partially different; each element in the dimensionality reduced text vector is the average of the elements in one of the N sub-vectors; and performing vector retrieval using the dimensionality reduced text vector.

This application belongs to the field of large model technology and discloses a method, device, and storage medium for evaluating the quality of youth activities based on LLM. The method includes: acquiring multiple text data containing natural language content and source identifiers for the same youth activity; encoding these data using a large language model to obtain text embedding representations; mapping these representations to a preset activity semantic dimension through an activity semantic anchoring layer to generate activity semantic anchor vectors; inputting the text embedding representations and anchor vectors into an inference layer; calculating the deviation between the hidden state and the anchor vector using a semantic stability constraint sub-network; adjusting the weight of the hidden state when the deviation exceeds a threshold; and finally generating an evaluation result based on the stability-constrained hidden state. This method effectively improves the stability, consistency, and verifiability of the evaluation results and is applicable to various youth activity quality evaluation scenarios.

The application discloses a kind of based on big model's exclusive team performance portrait generation and intelligent evaluation system, belong to artificial intelligence and performance management technical field.The system includes data acquisition module, index system management module, performance calculation engine module, portrait generation module, big model intelligent analysis module and visual display module.Index system adopts four-level structure, including six first-level indexes, twenty-four second-level indexes, fifty-eight third-level indexes and a number of fourth-level indexes.Performance calculation engine is based on difference ratio method to calculate index score, portrait generation module generates radar chart, star rating and ranking, and big model intelligent analysis module is based on big language model to carry out intelligent evaluation and ability analysis.The application realizes the scientificity, intelligentization and visualization of exclusive team performance management, and provides strong support for exclusive team construction decision.