72 results about "Context learning" patented technology

A test question recommendation method based on large language model adaptive multi-level evaluation constructs a multi-level architecture including semantic consistency verification, fine-grained fact alignment and dynamic cognitive evaluation, and comprises the following steps: constructing an initialized data stream and executing dynamic random sampling; calculating discrete semantic entropy by using NLI bidirectional implication clustering so as to quantify and eliminate illusion content with high uncertainty; under the RAG framework, the test questions are deconstructed into atomic statements, and a fact deviation is corrected by calculating a retrieval relevance vector and a logical implication consistency score; analyzing and screening low-quality texts based on multidimensional language features of syntax and logic; vector fusion is carried out on the generated intention and the cognitive portrait of the student, and a dynamic evaluation index system adaptive to a specific teaching scene is constructed in real time by utilizing context learning. The problems of accuracy and adaptability of automatic question setting are effectively solved, and intelligent closed-loop control from test question generation to cognitive alignment is realized.

Aspects of the disclosure are directed to methods, systems, and non-transitory computer readable media for automatically generating queries on a database from natural language text using in-context learning to leverage zero-shot and few-shot adaptation capabilities of large language models (LLMs). The methods, systems, and non-transitory computer readable media can consider database information, employ execution based consistency decoding, and employ a mixture of prompts and / or LLMs.

The invention discloses a sheet metal part manufacturability reasoning method based on space-semantic map alignment, and relates to the field of manufacturing-oriented design evaluation and industrial knowledge reasoning, and the method comprises the following steps: carrying out geometric analysis on a CAD geometric model of a sheet metal part to be evaluated; abstracting the geometric features and the topological / metric spatial relationship thereof into a computable spatial semantic graph; performing semantic analysis on the process specification described by a natural language, converting the process rule into formalized logic check expression by using a large language model through context learning, and generating an executable domain-specific language check script; executing the script on a spatial semantic graph, realizing deterministic reasoning through graph matching and attribute verification, and completing accurate mapping and violation detection of text rules and geometric features; and outputting an interpretable diagnosis result containing violation feature positioning, triggering rules and numerical evidence. In order to solve the problems that a process rule'natural language-geometric model 'has a semantic gap, a traditional rule system is poor in adaptability, and an end-to-end learning method is high in data dependence and cannot be explained, a new rule can be quickly adapted under the condition that a large amount of data does not need to be labeled and a model does not need to be retrained; the method can accurately identify the violation of the micro-size and spatial relationship, and has reasoning preciseness, interpretability and engineering availability.

The invention discloses a cross-organization emergency response process model extraction method based on a large language model, belongs to the field of process mining and natural language processing, and constructs a large language model general prompt framework containing emergency response process element definition and judgment criteria. Based on a general prompt framework, a prompt strategy fusing role playing and context learning is adopted to extract explicit entities such as emergency organizations and response tasks and relationships of the explicit entities. Inferring and extracting implicit entities and implicit relation elements on the basis of explicit elements by adopting a thinking chain fused prompt strategy; and finally, making a mapping rule to convert the extracted complete process elements into a logically coherent cross-organization emergency response process model, and presenting a result in a graphical manner. According to the method, domain data training is not needed, accurate flow element extraction can be achieved under the condition of few samples, the problem of implicit flow element missing is solved, and technical support is provided for emergency decision making.

The invention provides a large-model intelligent label synthesis and data automatic labeling integration method and system, and belongs to the technical field of label synthesis and data labeling, and the method comprises the steps: carrying out the semantic embedding and robust clustering of text data, and obtaining a stable cluster set; when new data is introduced, semantic consistency alignment of cross-round clustering results is realized through a confusion matrix matching strategy, and label drift is inhibited; maintaining an editable hierarchical label directed acyclic graph to support label system evolution; driving a large language model to generate a high-quality and interpretable cluster-level semantic tag based on the representative sample; carrying out automatic annotation and confidence evaluation by using large model context learning for clustering non-attribution or low-confidence samples; and propagating the cluster-level labels to the instances, and combining the cluster-level labels with an automatic labeling result to construct a full-process traceable label management mechanism. According to the method, the efficiency, quality and consistency of text labeling are improved, and powerful support is provided for large model training and intelligent data management.

The invention discloses a multi-modal deep forgery detection method and system based on multi-agent collaborative reasoning and context learning, and the method comprises the steps: carrying out the analysis of an image in an input image-text pair through a visual expert module, and generating a visual description corresponding to the content of the image; performing multiple rounds of interactive questions and answers on a text in the input image-text pair based on the visual description by utilizing a multi-agent module, and generating a dialogue history containing multiple rounds of questions and answers; the method comprises the following steps: searching Top-k examples similar to the semantics of a current image-text pair from a pre-constructed multi-modal knowledge base through a search module, and constructing a context example set; and by summarizing the expert module, fusing the dialogue history, the text and visual description of the current image-text pair and the context example set, outputting authenticity judgment and forgery type prediction of the to-be-detected image-text pair. According to the embodiment of the invention, the accuracy, generalization ability and result interpretability of multi-modal depth forgery detection can be improved.

The application discloses a kind of based on context learning's incremental user portrait method and system, belong to electric power system user portrait technical field, the method is: based on the small sample load data of to-be-tested incremental user construction incremental user portrait task pair;Obtain several current context task pairs, and incremental user portrait task pair and several current context task pairs are combined to form test sample;Based on test sample and pre-trained user portrait model, obtain incremental user portrait result;Pre-training process includes: obtaining the stock historical load data sequence set of several stock users;Based on stock historical load data sequence set, preset historical window length and preset future window length, construct the preset portrait model training data pool corresponding to several stock users;The pre-constructed basic portrait model is trained based on preset portrait model training data pool, and user portrait model is obtained, therefore, by implementing the application, small sample, high-precision portrait to incremental user can be realized.

The application discloses a method and device for efficient automatic power consumption optimization of chip physical design, solves the problem that the prior art cannot balance the time cost and optimization effect in design space exploration, realizes efficient exploration of high-dimensional continuous physical design parameters, and thus optimizes design parameters and reduces chip power consumption under the premise of guaranteeing design constraints; the method comprises the following steps: determining physical design parameters and boundaries based on a chip process node, sampling and constructing an initial data set; using a pre-trained large model to establish a small sample power consumption evaluation proxy model through context learning; parameter generation model iterative optimization: generating a candidate parameter combination, inputting the proxy model to obtain predicted power consumption, and updating model parameters until convergence according to the predicted power consumption; after training, the model generates a parameter combination to call an EDA tool to obtain real power consumption, so as to feed back the optimization model, and finally output a chip design with optimized power consumption.

This application relates to the field of cross-task shared optimization technology for code intelligence, and particularly to a method, apparatus, and electronic device for cross-task shared optimization of code intelligence. The method includes: acquiring a hierarchical context learning task and a code-natural language multimodal contrastive learning task; generating training data based on the hierarchical context learning task and the code-natural language multimodal contrastive learning task; training a pre-set pre-trained language model based on the training data; freezing the network parameters of the trained pre-trained language model; transferring the shared parameters of the pre-trained language model to the downstream code target task for parameter optimization; inputting the code to be tested into the optimized pre-trained language model; and outputting the optimization result of the code to be tested through the pre-trained language model. This solves the problems of weak cross-task transfer and generalization ability of pre-trained models and low parameter optimization efficiency in related technologies.

The application provides a large model intelligent label synthesis and data automatic labeling integrated method and system, and belongs to the technical field of label synthesis and data labeling. The method comprises the following steps: performing semantic embedding and robust clustering on text data to obtain a stable cluster set; when new data is introduced, the semantic consistency alignment of cross-round clustering results is realized through a confusion matrix matching strategy to suppress label drift; an editable hierarchical label directed acyclic graph is maintained to support label system evolution; a large language model is driven based on representative samples to generate high-quality and interpretable cluster-level semantic labels; for samples that are not attributed or have low confidence in clustering, automatic labeling and confidence evaluation are performed through context learning on the large model; the cluster-level labels are propagated to instances and combined with the automatic labeling results to construct a full-process traceable label management mechanism. The application improves the efficiency, quality and consistency of text labeling, and provides strong support for large model training and intelligent data management.

Systems and methods for optimized multi-modality processing with artificial intelligence models. Relevant page images can be extracted from a multi-modality index with a dynamic multi-modality processing (DMMP) system. Semantic and contextual relationship can be captured between a user query and multi-modality content of the relevant page images. The multi-modality content of the relevant page images can be converted to an instruction code based on the semantic and contextual relationship captured with a context learning module of the DMMP system. An issue identified by the DMMP system based on the instruction code that includes the user query can be corrected by generating a corrective action with the DMMP system by utilizing the instruction code.

The invention discloses a remote sensing small sample segmentation method and system based on multi-view context learning, and the method comprises the steps: carrying out the rotation data enhancement of a reference image, generating a plurality of reference images of different views, splicing the plurality of reference images of different views with a to-be-detected image, and obtaining a joint input image; uniformly distributed positive and negative prompt points are generated in each reference image area in the joint input image, the foreground area is marked as the positive prompt point, the background area is marked as the negative prompt point, coordinate information of the positive and negative prompt points is input into a prompt encoder, and sparse prompt feature embedding is obtained; inputting the joint input image into a pre-trained image encoder to extract features, and generating joint features; and jointly inputting the sparse prompt feature embedding and the joint feature into a mask decoder to generate a final segmentation mask. According to the method, a domain barrier between a natural image and a remote sensing scene is broken, and accurate segmentation of a target in any direction in a complex background is realized under the condition of few reference images.

Various embodiments provide methods performed by a user equipment (UE) including receiving sensor data associated with activity of a user from a plurality of inputs at a first-time instance, determining a state of the user at the first-time instance based on the sensor data, and identifying, by the processor, at least one attribute of the user or the activity of the user based on the sensor data and the state of the user collected at multiple time instances including the first-time instance. Identifying the at least one attribute may use an inference model to receive selected sensor data as an input and to provide an output classifying the state of the user at the multiple time instances as an attribute of the user.

This application provides a generative knowledge object extraction method and system based on context learning, relating to the field of generative artificial intelligence technology. The method includes: preprocessing the target text by encoding it using a first Transformer encoder in a dual-tower structure to obtain a target vector; traversing and matching the target vector in a pre-stored vector library to obtain a target matching vector, where the target matching vector corresponds to a target example; combining the target text and the target example to form a target structured input prompt; and under the constraint of the target structured input prompt, a generative knowledge object extraction model extracts and analyzes the target text to obtain the target extraction result. This application can solve the technical problem of poor accuracy in entity recognition results of existing generative entity recognition models, achieving the technical effect of improving the accuracy of generative named entity recognition models.

The application provides a low-illumination image enhancement method, adopts a scale-aware low-illumination image enhancement network, and enhances the low-illumination image in a self-adaptive mode, so that the problems of local overexposure and insufficient enhancement are avoided. The network is composed of a scale-aware feature extraction network, a scale-aware context extraction network and a residual up-sampling network. Firstly, aiming at the scale and shape diversity of the low-illumination area, a scale-aware transformer module is designed, which is aimed at fully mining the context information at different scales, so that the network has excellent local perception ability, thereby realizing high-quality illumination enhancement. Based on the module, a scale-aware feature extraction network is designed, which extracts feature information at different levels by using the module; at the same time, a scale-aware context learning network is designed, which further learns the global information of the image on the basis of the features with a large receptive field and rich semantics, so as to obtain the features containing the scale and context of the image. Finally, the features are processed through the residual up-sampling network, and gradually generate high-quality images with reasonable illumination and clear details.

A disclosed method facilitates AI-generation of a customized email per a methodology that significantly reduces the risk of the customized email including hallucinated facts or undesirable personal identity information (PII). The method includes identifying an email template and a recipient identifier that identifies a recipient of the customized email based on user inputs to an email application; mining contextual data stored in association with the recipient identifier; generating a large language model (LLM) prompt based on the email template and the contextual data; providing the LLM prompt as input to a trained large language model (LLM); receiving the customized email as an output from the LLM; and returning the customized email to the email application for display within a user interface.

The application discloses a method for generating an intelligent body recognizable structured language, and belongs to the technical field of intelligent body control. The method inputs a natural language task demand input by a current user and an external environment state parameter related to the natural language task demand collected by an external environment detection module into a large model, and finally obtains a structured text, so that an intelligent body device corresponding to the natural language task demand reads and executes a corresponding user demand task. The method for generating the intelligent body recognizable structured language fully utilizes a prompt word module splicing and context learning, adopts a language large model and an event extraction combination mode, generates an atomic event sequence and automatically modifies the atomic event sequence, effectively realizes intelligent task planning, and improves planning accuracy.

The invention discloses a dialogue relation perception cascade type multitask multi-party dialogue emotion recognition method fused with context learning. The method comprises the following steps: step 1) collecting public data sets of utterance relation analysis and dialogue emotion recognition; 2) training a dialogue utterance analysis model based on a dependency relationship by utilizing the utterance relationship analysis data, performing utterance structure analysis on the dialogue emotion data, and extracting a dependency relationship between utterances in the dialogue emotion data; 3) constructing instruction data by using the dialogue dependency relationship data, and performing dialogue dependency relationship task training on the large language model to obtain a pre-trained dialogue relationship perception model; the multi-party dialogue emotion recognition method comprises the steps of (1) obtaining a multi-party dialogue emotion recognition task instruction, (2) retrieving examples related to a current input dialogue from multiple dimensions, and providing a reference sample for context learning, and (3) adding the retrieved examples as context information into the multi-party dialogue emotion recognition task instruction, training a previous dialogue relation perception large language model, and performing multi-party dialogue emotion recognition by using the trained model.

The invention discloses a large language model equipment relation extraction method based on two-way retrieval and context learning, and relates to the technical field of relation extraction in natural language processing. The large language model equipment relation extraction method based on two-way retrieval and context learning comprises the following steps: S1, fine adjustment of an embedded model; s2, constructing a retrieval example library; s3, performing double-path retrieval and correlation sorting; and S4, prompt word construction and relation extraction. According to the large language model equipment relation extraction method based on two-way retrieval and context learning, global semantics and local instance relations are captured at the same time through two-way retrieval, three types of overlapping relations of Normal, SEO and EPO in the equipment field can be effectively recognized, and an F1 value is remarkably improved compared with an existing optimal baseline model GPT-RE; through comparison learning and a BGE-M3 model optimized by a domain adaptation layer, semantic information of technical terms and technical description in the field of equipment can be accurately captured, and the problem of insufficient domain adaptation of a general model is solved.

The invention provides a context learning method and system based on task decomposition difficulty and a medium, and the method comprises the steps: obtaining a retrieval data set and a query text, retrieving first k example pairs similar to the semantic of the query text from the retrieval data set, and constructing a context data set for query; obtaining an original task, and dividing the original task into a plurality of sub-tasks; constructing a tetrad comprising a training text, a context, a label space and a score, and setting the label space based on the maximum subtask width; the score is used for evaluating the label space of the current subtask, the label space with the minimum prediction error is selected as the next subtask to be divided, when no finer division can improve the prediction score, the process is ended, and the LLMs is directly predicted in the label space corresponding to the current subtask. By simplifying the original task, the SICL ensures that the capability of the LLM can exceed the task difficulty, thereby achieving good performance.

The invention discloses a context example selection method and system based on factorial hidden variables and common cause generation. The method comprises the following steps: constructing a factorial potential state space, performing refined modeling on task features, decoupling an abstract task concept into mutually independent potential factors, performing common cause and effect modeling, abandoning a binary cause and effect hypothesis, and regarding an input text and an output label as observation values jointly generated by the potential factors; a full-correlation decoupling penalty term is introduced to optimize a variational lower bound, a multi-head encoder is trained to identify and separate each independent factor, in a reasoning stage, demand distribution of an input text in each factor space is calculated and tested, a complementary example set capable of covering all target factors is retrieved and combined from a candidate pool, and the complementary example set can cover all target factors. Therefore, the problem that example selection is biased in a complex context learning task is effectively solved, the few-sample reasoning performance of a large language model is remarkably improved, accurate screening of examples is achieved, and the matching degree of the examples and a test input text is improved.

The invention discloses a scientific chart question and answer data generation method and system based on a multi-modal large model. The method comprises the following steps: designing a set of systematized question and answer data generation process; a large number of scientific charts are analyzed, and conditions, targets and operations are introduced to serve as three types of basic components for forming chart problems to form core elements of the chart problems; summarizing a corresponding candidate value set for each component based on deep observation of a large number of chart problem instances; and in the question generation process, sampling is performed from the candidate set of each component according to a preset difficulty level, and the components and the difficulty levels thereof are utilized to jointly guide the template to automatically generate the question through context learning. According to the method, a large-scale scientific chart question and answer data set can be efficiently and systematically constructed, and the performance and generalization ability of a multi-modal large model in chart understanding and question and answer tasks are remarkably improved.

A system and method of programmer-interpreter approach for large language model post-editing is described. A method includes receiving, by a generator in a text generation system, an input, translating, by the generator, the input into an initial output text using at least a set of examples retrieved by a function in the generator when performing in-context learning, iteratively refining, by a programmer and an interpreter in the text generation system, the initial output text or intermediate output text, wherein the programmer encodes domain task-specific knowledge and the interpreter facilitates domain generalization, iteratively improving a quality of the initial output text or the intermediate output text in low-resource cross-domain text generation tasks by exploiting encoding of domain task-specific knowledge by the interpreter and facilitation of domain generalization by the interpreter, and outputting, by the text generation system, output text based on the initial output text and the intermediate output text.

This invention discloses a method and system for integrated communication and sensing based on a large-scale intelligent reflector-assisted system. First, the invention characterizes the problems of channel estimation, beamforming design, and user location estimation during large-scale model training. Then, it designs a multi-task context learning mechanism, heterogeneous data reconstruction, and three network modules to organically incorporate three physically semantically distinct integrated communication and sensing tasks—channel estimation, beamforming design, and user location—into a unified model. Based on the uplink pilot signal sent by the user, the invention achieves the aforementioned three typical integrated communication and sensing tasks. Furthermore, this invention provides a differentiable reparameterized phase retrieval strategy for intelligent reflector phase shift, ensuring that the model output strictly meets hardware physical constraints. This overcomes the limitation of existing networks that can only handle single tasks, providing an efficient solution for large-scale intelligent reflector-assisted integrated communication and sensing.

The invention discloses a smoking target identification method based on context learning. The method comprises the following steps: S0, performing coarse screening by using skeleton extraction; the method comprises the following steps: S1, inputting an image into a pre-trained ResNet-101 network for feature extraction; s2, constructing a feature pyramid network, and fusing multi-scale features through a top-down path; s3, splicing to form a global feature map; s4, key features are enhanced through an scSE module; s5, candidate areas are generated and screened; s6, performing context alignment on the candidate areas; and S7, performing classification and regression. The false detection rate and the omission rate in a complex scene can be reduced, and high-precision real-time smoking behavior recognition can be realized as far as possible.

The present disclosure relates to machine learning techniques for In-Context-Learning (ICL) with pattern-based retrieval for the task of converting Natural Language (NL) to Structured Query Language (SQL). Aspects are directed towards acquiring a natural language utterance and a database schema, searching, using at least a portion of the natural language utterance as a key, a memory bank for one or more in-context examples that are relevant to the key, generating a prompt comprising the natural language query, the database schema, and the one or more in-context examples, transmitting the prompt to a first pretrained generative artificial intelligence model, receiving, from the first pretrained generative artificial intelligence model, a logical form corresponding to the natural language utterance based at least in part on the prompt, executing the logical form on a database to obtain a query result, and providing the query result to a user.

The application discloses a kind of based on skeleton decomposition and sequence completion's time sequence damaged fine-grained action recognition method, through context learning module, the query pair of hint in priori skeleton database is used to the sequence to be repaired, restore missing frame in time sequence damaged skeleton sequence, generate substantially complete skeleton sequence;According to human structure, joint point is divided into multiple semantic regions, and according to the motion intensity of each region, it is divided into dynamic and static region, respectively, different spatiotemporal disturbance and reorganization are applied to strong, weak, and enhanced dynamic sequence and static sequence are constructed;Based on Lagrange dynamics principle, acceleration sequence is obtained by MLP modeling and fusion;Finally, cross fusion features are completed by graph convolution network Action classification.The application effectively improves the robustness of the model under the condition of incomplete time sequence, enhances the discriminant ability to subtle action difference, and achieves higher recognition accuracy than existing methods on public data sets.

A method includes receiving a natural language query and determining that the natural language query is requesting generation of domain specific language (DSL) code. Based on determining that the natural language query is requesting generation of DSL code, the method includes retrieving a subset of documents from a search index including a set of documents based on the natural language query. Each document of the set of documents includes a respective portion of a DSL specification paired with a respective natural language context. The method includes generating a prompt based on the natural language query and the subset of documents. Using a large language model (LLM), the method includes generating the DSL code based on the prompt.

A vision inspection method using in-context learning, and a system therefor, according to one embodiment of the present disclosure, relate to a method and a system for performing, by using sequence model-based in-context learning, vision inspection on a query image by controlling an operation of a model through a prompt including a small number of images and label examples without updating model parameters.