187 results about "Context learning" patented technology

The invention discloses a civil administration service question and answer method based on a large model and knowledge graph retrieval enhancement, and the method comprises the steps: S10, inputting a user question, and carrying out the question analysis and entity recognition; s20, the problem complexity is judged, if the problem is a single-hop problem, knowledge graph single-hop retrieval is carried out, and if the problem is a multi-hop problem, knowledge graph multi-hop retrieval is carried out; s30, performing semantic matching sorting to generate sub-answers; and S40, based on the sub-answers and the user question, performing synthesis to generate a final answer. According to the method, the structured knowledge of the knowledge graph and the natural language processing capability of the large language model are fused; a question decomposition module is used to enhance the interpretability of multi-hop information retrieval and answers; and using contextual learning (ICL) and thinking chain (CoT) prompts to generate an individually processed explicit inference chain to improve authenticity; the defects of traditional civil administration service questions and answers in the aspects of knowledge accuracy, reasoning ability and interpretability are overcome.

One or more systems, devices, computer program products and / or computer-implemented methods of use provided herein relate to enhancing LLMs using in-context learning and online knowledge. The computer-implemented system can comprise a memory that can store computer executable components. The computer-implemented system can further comprise a processor that can execute the computer executable components stored in the memory, wherein the computer executable components can comprise a semantic retrieval component that can extract information from an online source, according to a query, to generate an in-context learning input utilized by an LLM for responding to the query.

The invention discloses a sports video commentary generation method and system based on a multi-modal large language model, and the method comprises the steps: obtaining a multi-modal data set which comprises a sports video, and an audio and a commentary text corresponding to the sports video; constructing a multi-modal large language model, encoding the sports video, the audio and the explanation text so as to project corresponding video frames, audio waveforms and metadata to a shared embedding space, and determining a multi-modal embedding vector; a multi-modal clustering memory unit is set, multi-modal embedded vectors are grouped, and feature alignment between modals is optimized through comparative learning and information entropy regularization; based on a retrieval enhanced context learning mechanism, a historical instance is retrieved through sparse regularization distance measurement to serve as reference input of a current input multi-mode embedded vector; and jointly inputting the current multi-modal embedded vector and the reference input into the multi-modal large language model to obtain a sports video commentary. According to the method and the device, the problems of insufficient multi-modal information integration and insufficient context utilization are solved.

The invention relates to the technical field of large language models, in particular to a complex multi-field guided teaching question and answer generation method, device and equipment and a computer storage medium. According to the complex multi-field guided teaching question and answer generation method, aiming at the problem that the model is single in application field, the corresponding multi-round question and answer pair data set is processed by using a multi-field data source, and the model is finely adjusted based on the multi-round question and answer pair data set, so that the problem that the related field is single is solved, and the applicability and practicability of the teaching question and answer model are enhanced; aiming at the problem that the model lacks guiding ability, the input of the teaching question and answer model is further processed through thinking chain-based question extraction and disassembly and prompt engineering and context learning mechanism-based processing, so that the model question guiding ability is improved; in order to solve the problem that teaching effect evaluation indexes are single, the invention provides a new evaluation system, and a plurality of evaluation indexes can more comprehensively and accurately evaluate and optimize model teaching effects.

Methods and systems for prompting a Large Language Model (LLM) with a set of text data outside pre-inference trained categories and a test prompt for an initial parameter which has a known ground truth, calculating an uncertainty of an LLM's output, selecting another LLM model parameter and calculating the total uncertainty of the LLM's output with the other LLM model parameter. The methods and systems further include prompting the LLM with another test prompt, with the initial LLM parameter and the other LLM parameter, and calculating the total uncertainty of the LLM's output for initial LLM model parameter and the other LLM model parameter, decomposing the total uncertainty of the LLM into Aleatoric Uncertainty (AU) and Epistemic Uncertainty (EU) components, and rating the total uncertainty of the LLM, using the decomposed total uncertainty as a metric.

The invention discloses a recommendation system deviation correction method fusing big language model world knowledge, which comprises the following steps: acquiring an open source data set of a recommendation system, and generating a training set; guiding a large language model to generate description information of an article and preference reasoning of a user through a training set in combination with a context learning technology; encoding the description information of the article and the preference reasoning of the user through a pre-trained text encoder to generate a first encoding vector, and encoding the IDs of the user and the article to generate a second encoding vector; inputting the first coding vector and the second coding vector into a multi-mode expert network module to obtain final representations of the user and the article; inputting the final representations of the user and the article into a prediction layer to obtain a prediction value, and optimizing a multi-modal expert network module through a cross entropy loss function; and performing recommendation system correction through the optimized multi-mode expert network module. The inherent deviation in the user behavior data is eliminated, the real preference of the user is captured, and recommendation system deviation correction is completed.

An example operation may include one or more of storing a table comprising a plurality of columns corresponding to a plurality of attributes and a plurality of rows corresponding to a plurality of records, receiving a target record of a task of an artificial intelligence (AI) model, converting the plurality of records into a plurality of embeddings in multi-dimensional vector space, converting the target record into a target embedding in the multi-dimensional vector space, identifying a subset of records from among the plurality of records that are nearest to the target record in content based on distances between embeddings of the subset of records and the target embedding within the multi-dimensional vector space, and executing the AI model on the subset of records to generate a predicted output with respect to the task.

The invention discloses an image annotation method, system and device for spacecraft target type and attribute classification and a medium, and belongs to the technical field of image recognition. The method comprises the following steps: constructing a multi-dimensional attribute model based on key elements of a spacecraft image; based on the multi-dimensional attribute model, constructing a reference prompt template, obtaining a preliminary output result through the reference prompt template, and carrying out optimization processing on the reference prompt template and the preliminary output result by adopting few-sample learning to obtain a first output result; performing output supervision on the thinking process of the first output result in combination with the thinking chain, and obtaining a second output result through an optimization prompt template; and according to the first output result, the second output result and the plurality of output results, generating an output result data set, and obtaining an image annotation result. According to the method, through a semi-supervised labeling method combining multi-attribute modeling and context learning, the accuracy and robustness of fine-grained labeling of the spacecraft image in a few-sample scene are improved by utilizing a generative large model.

The invention relates to a large language model dynamic routing method and device based on context learning model characterization and a readable storage medium. Query is embedded and mapped to a language model input space by using a projection model, semantic alignment is realized, a representative evaluation set covering multi-dimensional capability is automatically screened from a benchmark question bank, and the representative evaluation set is used for evaluating the multi-dimensional capability. Performance characteristics of the model on an evaluation set are efficiently obtained at a time, and high-quality context model capability representation is formed; then real-time query embedding and context model capability representation are combined, a lightweight routing language model is used for supervised learning, so that the fine-grained model distinguishing capability is achieved, an increment embedding updating mechanism is designed, and when a new model is accessed or an old model is upgraded, cold start can be rapidly completed only through a very small number of fixed questions, so that the efficiency is improved. The calculation and maintenance cost is greatly reduced, and the accuracy, real-time performance and flexible expansibility of model routing are effectively improved.

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.

The method comprises the following steps: constructing an intelligent research and judgment model, carrying out environment modeling on the intelligent research and judgment model, constructing a network attack path according to the intelligent research and judgment model, and identifying a network attack by using a known signature of the network attack or a rule-based detection technology; classifying and marking the data of the network security equipment by using an SVM classification algorithm, training an intelligent study and judgment model, establishing a threat identification strategy of the network attack, and identifying the features of the network attack; performing context learning and state prediction by using the intelligent study and judgment model, and optimizing a threat identification strategy of network attacks; a data field of a normal HTTP request is slightly disturbed, an adversarial sample is generated by using a threat of a known network attack, adversarial training is performed in combination with a forged HTTP request, and a deceptive network attack is identified. According to the method, cross validation and correlation analysis of high-quality and multi-source threat intelligence are realized, and the attack defense capability is improved.

The invention provides a medical intelligent causal decision-making and scheduling system based on an improved large language model, and the system comprises a pre-training language model module which is used for processing unstructured medical text data, and extracting a text feature vector and a medical entity relationship; the knowledge graph enhancement module is used for constructing a medical knowledge graph based on the medical entity relationship, and reasoning to generate a knowledge enhancement feature vector; the multi-modal feature fusion module is used for generating a fusion feature vector based on the text feature vector and pre-stored structured data; the dynamic time sequence modeling module is used for generating a time sequence prediction vector based on the fusion feature vector and the knowledge enhancement feature vector; the self-adaptive scheduling optimization module is used for making a medical scheduling decision based on the time sequence prediction vector and generating a scheduling scheme; and the scheduling calibration module is used for matching similar case features for the scheduling scheme through a context learning mechanism based on the knowledge enhancement feature vector and calibrating the similar case features to generate an optimized scheduling scheme. And medical resources and scheduling can be allocated more accurately.

The invention discloses a medical insurance field document intelligent question-answering system and method based on an AI large model, and belongs to the technical field of artificial intelligence. The technical problems that an existing medical insurance question-answering system depends on manual configuration and is poor in flexibility and low in intelligent level are solved. According to the technical scheme, the system comprises a document processing module for analyzing and segmenting original documents such as medical insurance policy documents, vectorizing the original documents and storing the original documents into a vector database; and the knowledge question and answer module is used for analyzing user questions by using an AI large model, driving AIAgent to retrieve related knowledge in a vector database, performing reasoning in combination with injected professional knowledge such as business dictionaries and analysis logic, and generating accurate and traceable answers. Through context learning, logical reasoning and automatic generation ability of a large model, tedious intention model training and manual QA configuration are not needed, the intelligent level, adaptability and user experience of a medical insurance question-answering system are remarkably improved, and the operation and maintenance cost is reduced.

The embodiment of the invention relates to the technical field of large model data analysis, and discloses a large model evolution multi-agent investment decision matrix system. The investment Transform module is used for acquiring financial data and performing feature extraction and feature fusion on the financial data in combination with a context learning enhancement function and a financial domain knowledge enhancement function; the interpretability analysis module is used for acquiring the feature representation in the investment Transform module and carrying out feature relation mapping and interpretability analysis; the multi-agent decision-making module is used for carrying out collaborative decision-making through multiple agents; the optimization module is used for providing interpretation optimization for the interpretability analysis module and providing strategy optimization for the multi-agent decision-making module; and the knowledge base module is used for providing an investment template for the investment Transformer module, providing causal relationship knowledge for the interpretability analysis module and providing rule constraints for the multi-agent decision-making module. The technical problem of investment decision analysis can be at least solved.

Aspects provide in-context learning with calibration for a foundation model. A validation prompt including one or more demonstration samples and an evaluation example is received and an output probability generated using the foundation model and the validation prompt. A calibration loss is computed based on the output probability and a set of calibration parameters. The set of calibration parameters is updated based on the calibration loss using an optimization algorithm. An inferencing operation is performed using the updated calibration parameters.

The invention discloses a knowledge graph inference method based on prior knowledge enhancement, and mainly solves the problem of inaccurate reward caused by insufficient knowledge graph prior knowledge in the existing knowledge graph inference method based on reinforcement learning, thereby optimizing a model more effectively. The method comprises the following steps: reasoning path searching based on reinforcement learning; calculating an answer correctness reward; logic rationality reward calculation; a reward enhancement strategy based on path importance; performing logic rationality reward enhancement based on large language model context learning; and optimizing the model. According to the method, on the basis of an existing knowledge graph reasoning method based on reinforcement learning, huge internal knowledge of a large language model is fused into reward calculation in an efficient and information loss resistant mode, and the problem that rewards are inaccurate due to insufficient prior knowledge of a knowledge graph is solved. The performance of the knowledge graph reasoning method based on priori knowledge enhancement is remarkably improved compared with that of an existing method.

The invention discloses a small sample unified granularity relation extraction method based on a large language model, which comprises the following steps of: firstly, giving a specific task description as a part of an input context of the large language model; thirdly, giving an analogous example to the large language model as context demonstration; in order to better prompt the position information of the entity of the large language model in the context, performing entity enhancement on the context input into the large language model; and finally, a mode for serializing the relation triad is defined, and thinking chain reasoning information is fused in the mode, so that a large language model can be helped to perform relation extraction by utilizing thinking chain prompts. According to the method, context learning, thinking chain and entity enhancement technologies are introduced for unified granularity relation extraction tasks including a sentence level, a document level and a cross-document level, the powerful reasoning ability of a large language model is fully played, and the effectiveness of the model in the unified granularity relation extraction task, especially in a small sample scene, is improved.

Systems and methods are disclosed that generate dense blob representations such as blob parameters and blob descriptions, and use the dense blob representations to generate images. For example, embodiments of the present disclosure may decompose a scene into visual primitives (e.g., dense blob representations) and based on the blob representations, embodiments of the present disclosure develop a blob-grounded text-to-image diffusion model (BlobGEN) for compositional generation. For example, in some embodiments, a new masked cross-attention module may be introduced to disentangle the fusion between blob representations and visual features. In some embodiments, to leverage the compositionality of large language models (LLMs), a new in-context learning approach may be introduced to generate blob representations from text prompts.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for performing in-context learning using a generative neural network. In one aspect, a method comprises obtaining a plurality of demonstration examples for a task; obtaining a respective difficulty classification for each of the demonstration examples; generating a context input that includes one or more instances of at least a subset of the demonstration examples, the generating comprising, for each of the demonstration examples, determining how many instances of the demonstration example to include in the context input based on the respective difficulty classification for the demonstration example; receiving a new input for the task; and processing an input that includes the context input and the new input using the first generative neural network to generate a new output.

The invention provides a named entity recognition method based on multi-modal large model fine-grained knowledge generation, and belongs to the technical field of information extraction in natural language processing. According to the method, a multi-modal large model MLLM and a large language model LLM are combined to jointly generate fine-grained auxiliary knowledge, the LLM uses a context learning mode to guide the LLM to generate auxiliary knowledge related to a sample, and the MLLM uses manual annotation data to perform fine tuning on the LLM, so that the MLLM outputs the related auxiliary knowledge according to input text and image information; and combining the obtained auxiliary knowledge with the original text, entering a downstream sequence labeling model for training and reasoning, and completing named entity recognition. According to the method, world knowledge and multi-modal information which are beneficial to information extraction are taken into consideration, the performance of named entity recognition of the downstream sequence marking model is improved, and the entities in the text can be recognized more accurately.

The invention belongs to the field of intelligent workshop scheduling, and particularly discloses a large model enabled workshop scheduling end-to-end self-decision engine, method and device. According to the method, the large model is positioned as a core decision maker instead of a traditional auxiliary tool, so that the powerful capabilities of the large model in the aspects of natural language understanding, context learning, complex logical reasoning and content generation are fully utilized; according to workshop real-time state information, a dynamically changing workpiece pool, a scheduling target input by a user through a natural language and emergency description, an autonomous and end-to-end scheduling decision can be directly carried out, so that the rapid response capability and intelligent processing level of a scheduling system to dynamic events are improved, and the scheduling efficiency is improved. And the application threshold is reduced through natural language interaction, the dependence on explicit programming and rule customization is reduced, a low-code development paradigm is developed, and the method has a far-sighted decision potential exceeding a traditional heuristic rule.

The invention discloses a large model fine tuning method and system based on a suffix prompt learning method and ensemble learning, and the method comprises the steps: carrying out the preprocessing of an existing data set, obtaining K subsets through the replacement sampling, enabling each subset to account for a specified percentage of the total, and constructing an artificial prompt; according to the teacher model and the student model, a predicted target word classification score is obtained through suffix prompt learning, through probability normalization, teacher knowledge is distilled to the student model by using a KL divergence loss function, and a prompt vector is initialized; an input task vector is obtained through context learning, and knowledge of the input task vector is distilled to a student model prediction vector by using a mean square error loss function; constructing a mixed loss function, and optimizing knowledge distillation and task feature learning; performing end-to-end optimization on the initialized prompt vector by using suffix prompt in combination with a negative log-likelihood loss function; during reasoning, after the optimized prompt vectors are spliced to the input vectors, suffix prompts share the input vector key value cache to generate output. According to the method, the adaptability and reasoning performance of the model are remarkably improved.

The invention discloses a method, a system and a medium for understanding and identifying network moduli based on a multi-modal large model, and relates to the technical field of artificial intelligence, a to-be-processed multi-modal task is input into a trained context example selection model, and a context prompt example used for improving the context learning ability of the multi-modal large model is output; the training process of the context example selection model is as follows: performing feature extraction and modal alignment on a multi-modal task sample to obtain cross-modal joint embedding representation; performing coarse-grained sorting on a preset candidate sample library by using cross-modal joint embedding representation, and constructing an initial candidate set; performing fine-grained sorting on the initial candidate set, constructing a nested prompt template based on the obtained knowledge extraction example, and guiding a multi-modal large model to execute a classification reasoning task by taking the nested prompt template as a context prompt example; according to the method, the harmful content recognition precision and the cross-modal reasoning capability of the multi-modal large model in a low-resource scene are remarkably improved.

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 provides a large model context learning machine translation method based on word granularity alignment, which relates to the field of natural language processing, and comprises the following steps: an external knowledge assisting stage: performing multi-level retrieval matching on a source text word alignment set; a large model translation stage: obtaining a large model translation set, and taking the large model translation set as one of candidate translations; in the post-selection stage, the obtained source text word alignment set, the obtained external dictionary word alignment set, the obtained entity library alignment set, the obtained embedding representation of the auxiliary translation set and the obtained large model translation set are subjected to similarity calculation scoring for multiple times, and screening is conducted according to similarity scores to obtain a candidate word alignment set; a prompt template is designed according to a task, a source text word alignment set and a candidate word alignment set are put into the prompt template, and a highly-aligned external word alignment set is used for large model context learning to generate an optimal translation result; according to the method, various translation errors of a large model in a low-resource environment are relieved.

The invention discloses a reinforcement learning coach-driven context learning general motion control method and system, and the method comprises the steps: receiving the structure parameters of a plurality of predefined robots, carrying out the random sampling in a preset parameter space, and generating a plurality of robots of different structures; target tasks are defined, and reinforcement learning coaches corresponding to the target tasks are trained based on robots of different structures; in the simulation environment, guiding the robot to execute tasks by using a reinforcement learning coach, and recording a task execution track corresponding to each target task; extracting a context sequence with a fixed length from the task execution trajectory by using a general context learning framework, and performing cross-task meta-training by taking maximization of expected accumulated rewards under each target task as a training target so as to optimize context learning model parameters and obtain a pre-trained robot control model; a pre-trained robot control model; and obtaining a target control action which is output by the pre-trained robot control model and corresponds to the current state.

The invention provides a knowledge editing method based on multilingual semantic retrieval, and belongs to the technical field of natural language process.The method achieves cross-lingual knowledge updating through the two stages of multilingual knowledge retrieval and context editing and comprises the steps that firstly, a multilingual retrieval model based on XLM-R is used for mapping queries and knowledge base entries to a shared semantic space, and then the shared semantic space is used for conducting context editing; semantic correlation is judged through a classifier, and target language knowledge is retrieved; and splicing the retrieval result and the query in a zero sample or small sample mode to generate a prompt template, and inputting the prompt template into a large language model to complete editing. The method is characterized in that monolingual limitation is broken through, collaborative updating of 12 languages is supported, a retrieval-editing decoupling architecture is adopted to achieve model irrelevant adaptation, and retrieval and context learning are fused to improve editing accuracy. The method can efficiently correct multi-language large model fact errors, is suitable for scenes such as search engines and intelligent customer services, avoids the cost of full-model retraining, and remarkably improves the efficiency and accuracy of cross-language knowledge updating.

The present invention relates to an electric power distribution network fault processing device applied to the field of power distribution technology, aiming to improve the fault prediction, diagnosis and processing capabilities of the power system. The device includes a pre-training module, a neural network architecture module, a data acquisition layer, an edge computing layer, a core processing layer, a context learning module, an inference processing module, an output and interaction layer, and a continuous learning and improvement layer. The core of the present invention lies in the use of large-scale pre-trained Transformer models, combined with power system expertise, to achieve in-depth understanding and analysis of complex power grid states. Through multi-source data fusion, edge computing and center collaborative processing, the device can monitor the power grid state in real time, accurately predict potential faults, and provide intelligent diagnosis and repair suggestions. Self-supervised learning and federated learning technologies are used for model pre-training and optimization; reinforcement learning is used to optimize fault response strategies; the model's interpretability is achieved, and decision-making transparency is improved.

State of art techniques using moderate sized Language Models (LMs) for text classification need fine-tuning or in-context learning. A method and system providing a two-step classification using moderate-sized (#params≤2.7B) causal LM (Gen AI) is disclosed. Firstly, for a text instance to be classified, a set of perplexity and log-likelihood based features are obtained from an LM. Further, a light-weight classifier is trained in the second step to predict the final label. The system enables a new way of exploiting the available labelled instances, in addition to the existing ways like fine-tuning LMs or in-context learning. It neither needs any parameter updates in LMs like fine-tuning nor it is restricted by the number of training examples to be provided in the prompt like in-context learning. The key advantages of the disclosed system are explainability through most suitable key phrases and its applicability in resource poor environment.

Provided in the embodiments of the present description are a code error analysis method and apparatus, and a log scanning rule generation method and apparatus. The log scanning rule generation method comprises: acquiring a program code to be analyzed which corresponds to a code error analysis task; determining an analysis task data group on the basis of said program code, wherein the analysis task data group comprises said program code, and a first error code sample and a corresponding log scanning rule sample; and inputting the analysis task data group into an error analysis model, such that the error analysis model generates a log scanning rule, wherein the error analysis model is used for performing context learning on the basis of the first error code sample and the corresponding log scanning rule sample, and generating the corresponding log scanning rule for said program code, and the log scanning rule is used for performing error scanning on a program log. Therefore, the efficiency of error analysis for a program code is improved, and the development efficiency is improved.