121 results about "Contextual reasoning" patented technology

An AI-based Cybersecurity System and Method enable real-time detection, analysis, and mitigation of cyber threats within computing networks using adaptive artificial intelligence. The system continuously monitors network traffic, extracts behavioral and contextual attributes, and applies deep learning-based inference to identify anomalous activities indicating security breaches. The method integrates several computational units, including a network monitoring unit, feature extraction unit, artificial intelligence processor, contextual reasoning processor, and decision synthesis unit, to compute a composite risk index quantifying threat likelihood and severity. A classification processor categorizes detected threats into types such as ransomware, phishing, or unauthorized access, while a mitigation control processor initiates automated response actions to isolate compromised nodes and restore network integrity. An adaptive learning processor updates AI models using feedback from confirmed incidents. This provides a scalable, self-evolving cybersecurity framework that minimizes human intervention and enhances resilience against dynamic and zero-day threats.

The invention provides a front-end automatic testing method of a large language model based on LLM, and belongs to the technical field of front-end automatic testing. The method comprises five core steps of test intention recognition and semantic analysis, page structure analysis and target element positioning, automatic generation of a test operation sequence and assertion logic, test script execution and dynamic regression verification, and test feedback analysis and continuous training. The strong semantic understanding and context reasoning capabilities of the LLM are utilized to perform semantic analysis and simulation on a front-end page structure, a user behavior process and business logic, so that automatic generation and maintenance of a test script are realized. According to the method, the test coverage rate and the generation efficiency are greatly improved, the maintenance cost is remarkably reduced, the readability and the intelligence of the test case are enhanced, and the method is suitable for the modern Web front-end application test of rapid iteration.

Aiming at the problems of insufficient language understanding depth, weak personalized dialogue generation ability, lack of continuous learning and long-term user state modeling and the like in the current emotion recognition and psychological intervention technology, the invention provides a user emotion recognition and psychological intervention method combined with a large language model (LLM). According to the method, the potential emotional state is identified by analyzing free text information input by a user by utilizing the powerful capabilities of a large language model in the aspects of natural language understanding, emotional modeling and text generation; constructing a multi-round dialogue context, and reasoning a psychological change trend of the user; in combination with a psychological knowledge base, personalized and mild psychological intervention dialogue content with a dredging effect is generated. The system supports recognition and classification of various emotional states such as depression, anxiety and alonity, is suitable for various interaction scenes (such as APPs, webpages and social robots), and can greatly improve the precision of emotion recognition and the timeliness and effectiveness of psychological intervention. The emotion recognition and psychological intervention method based on the large language model provides solid technical support for constructing an intelligent, continuous and personalized psychological health management system, and has wide application prospects and profound social significance.

A real-time anomaly detection system for high-volume managed file transfers (MFT), consisting of: a secure, hardware-accelerated monitoring unit configured to interface with a managed file transfer server and intercept file transfer session data at wire-speed; a metadata extraction engine embedded in the hardware-accelerated unit, the metadata extraction engine configured to analyze protocol-specific session attributes, including, but not limited to, file size, transfer duration, encryption status, source and destination endpoints, transfer frequency, and payload entropy; a contextual AI inference engine communicatively coupled to the metadata extraction engine, the AI inference engine comprising a deep learning model trained on labeled historical MFT activity logs to detect contextual deviations from normative behavior; a federated learning architecture with a plurality of edge nodes, each hosting a local anomaly detection model trained on localized transmission metadata and configured to synchronize with a central aggregator using differentially private gradient updates; an Explainable AI (XAI) subsystem integrated into and configured to generate human-readable anomaly justifications, feature importance maps, and threat categorization labels; and a policy orchestration module configured to dynamically execute pre-configured or AI-based security responses, where the security responses include selective session termination, quarantining of transferred files, generation of alerts, or redirection of MFT workflows.

The invention relates to the technical field of artificial intelligence, can be applied to the fields of financial science and technology and medical health, and discloses a reasoning optimization method, device, equipment and medium for a multi-modal large language model.The method comprises the steps that an input long context sequence is obtained, and key value projection is conducted on the long context sequence to generate an initial key value cache; for each attention layer of the multi-modal large language model, calculating an attention matrix of the attention layer according to the vector dimension of the long context sequence and the initial key value cache; calculating a cross-modal attention entropy according to the attention matrix, and determining a cache size of an attention layer according to the cross-modal attention entropy; optimizing the initial key value cache based on a cumulative attention scoring mechanism and a window strategy to obtain a target key value cache; and reasoning the long context sequence according to the cache size and the target key value cache to generate a long context reasoning result. And the reasoning efficiency and the reasoning accuracy are improved.

The invention provides a task disassembly and multi-agent arrangement execution system and method based on a large language model, and belongs to the technical field of computers, and the system comprises an instruction analysis module, a DAG construction module, a scheduling execution module and a cache optimization module. Analysis is carried out according to the dependency relationship between the tasks, a task execution DAG is automatically constructed, and concurrent calling is carried out on the sub-modules without dependency; by caching an authentication result, a context reasoning result and the like, a universal module is executed in advance, the result is reused, and repeated calculation is reduced; the maintainability and the expandability of the system are improved through graph structure visualization and node element information injection; and performing context analysis and scheduling optimization in combination with the reasoning ability of the language model to realize an intelligent decision execution path.

The invention discloses a multi-agent task arrangement method and system, and relates to the technical field of artificial intelligence. In the method, firstly, a task demand is obtained, key information and intention of the task demand are extracted, and task semantics are obtained; secondly, based on the task semantics and the business standard process template, obtaining a business standard process template matched with the task demand; then, according to the matched business standard process template and task semantics, a task context is constructed, and a domain specific language DSL is generated through reasoning according to the task context; thirdly, the DSL is analyzed, a task dependency graph is constructed according to the DSL analysis result, and the task dependency graph is converted into BPMN process model data; and finally, sub-task scheduling and execution of the sub-agents are carried out according to the BPMN process model data. According to the method provided by the invention, the dependence on professional developers or process engineers can be reduced, the response and planning time of complex tasks can be shortened, the result predictability can be improved, and the reasoning cost can be obviously reduced.

A system for context-sensitive orchestration of autonomous agents in cloud platforms, consisting of: a hardware-based orchestration device configured for integration into a distributed cloud infrastructure; a context inference engine within the orchestration device, wherein the context inference engine is configured to receive and aggregate real-time telemetry data from a variety of distributed nodes, including at least one system-level parameter, at least one application-level parameter, and at least one environment parameter; a semantic inference module within the context inference engine, configured to generate a context-related state representation by correlating the parameters using a knowledge graph-based model of interdependencies; an optimization unit for machine learning within the orchestration device, which is communicatively connected to the context inference engine and is configured to predict resource requirements and operational states using reinforcement learning models trained on historical and real-time data streams; a policy-driven orchestration controller configured to translate the contextual state representation into actionable orchestration decisions by applying dynamic orchestration policies stored in a domain-specific policy repository; and a distributed agent interaction bus configured to delegate orchestration decisions to a variety of autonomous agents deployed on the cloud platform.

The invention discloses an algorithm method for large-model long context reasoning, which relates to the technical field of large language models and comprises the following steps of: dividing an input long text sequence into a plurality of initial text blocks; a semantic abstract is generated based on the initial text blocks, clustering analysis is conducted on the semantic abstract, the initial text blocks with similar semantics are combined into semantic hyperblocks, and a context organization structure with semantic representativeness is formed; the method comprises the following steps: generating key value cache data of each token in a large model preprocessing stage, grouping the key value cache data by taking semantic hyperblocks as logic boundaries, and establishing a mapping table for recording storage positions and states of the semantic hyperblocks; on the basis of semantic hyperblocks, semantic representative vectors of the semantic hyperblocks are combined, correlation scores between query vectors and the semantic hyperblocks are calculated, a block importance prediction model learns on the basis of context dependency features, high-order semantic prior is provided for subsequent attention screening, and key information omission caused by position offset is avoided.

The invention provides a context compression method based on query guidance. The context compression method comprises the following steps: filling query information and a context to be compressed into a dialogue template; respectively calculating first attention numerical values of a plurality of lexical elements in the context of the dialogue template; performing attention reconstruction on the context according to the first attention numerical values corresponding to the plurality of lexical elements; and compressing the context after the attention reconstruction to be within a preset target compression ratio. The invention further provides a context compression system based on query guidance, a storage medium and electronic equipment. Therefore, key information can be more efficiently reserved, a shorter context is input to the generative model, the reasoning efficiency of the generative model in a long context reasoning scene is improved, and the reasoning cost is reduced.

The invention provides a large language model long text reasoning acceleration method and device based on a speculative key value cache sparse technology, a medium, a terminal and a program product, the method is applied to electronic equipment comprising a GPU and a CPU, and the method comprises the steps that an index list of most important historical tokens is generated based on a distillation language model according to an obtained context sequence; comparing the index list generated at the current moment with the index list at the previous moment, and calculating to obtain a difference set part; asynchronously prefetching the key value cache of the difference set part from the CPU to the GPU; according to the asynchronously prefetched key value cache, performing parallel execution based on a large language model to generate a new token; obtaining a new sequence length according to the generated new token, and judging whether the new sequence length exceeds a preset threshold value or not; and if the threshold value is exceeded, executing unloading operation. According to the method, the performance and the stability of processing long text reasoning by the large language model can be improved, and key value cache optimization in the long context reasoning process is ensured to be always effective.

The invention relates to a request result generation method and device, a large model reasoning architecture, a vector database, equipment, a storage medium and a program product. The method comprises the following steps: unloading key value cache data obtained in a pre-filling stage to a vector database decoupled from a large language model for storage, and executing context vector retrieval and attention score calculation in a decoding stage in a text generation task by adopting the vector database decoupled from the large language model, therefore, the large language model can quickly reuse the target context vector obtained by the vector database and the attention score of the target context vector to carry out text generation reasoning, and a request result corresponding to the user request data is generated. By adopting the method, the data processing amount of the large language model can be greatly reduced, and the long context reasoning cost of the large language model is reduced.

The invention provides a fault question-answering method and device for a ship electric propulsion system and fault diagnosis equipment, and belongs to the field of ship fault diagnos.The method comprises the steps that a fault ontology model of the ship electric propulsion system is built, and a fault knowledge graph is built according to the fault ontology model and fault text data of the ship electric propulsion system; performing entity node matching on a to-be-diagnosed propulsion system fault problem and the fault knowledge graph to obtain a core entity set, and performing entity relationship expansion and redundant path pruning on the core entity set to obtain a knowledge graph sub-graph; and according to the to-be-diagnosed propulsion system fault problem and the knowledge graph subgraph, performing large language model path reasoning exploration and question and answer output to obtain a fault diagnosis result. According to the method, reasoning path text data support is provided for fault diagnosis by constructing the knowledge graph sub-graph, the knowledge illusion problem caused by wrong entity positioning in the problem reasoning process is avoided, and the context reasoning ability of fault questions and answers is effectively improved.

The invention discloses a version knowledge graph reasoning method and system based on large language model enhancement, and relates to the technical field of dynamic knowledge graphs, and the method comprises the steps: employing a semantic drift detection and compensation mechanism, comparing context coding differences of same entities of different versions in an initial knowledge graph, recognizing drift, and dynamically adjusting entity embedding vectors, outputting a compensation update atlas; an LLM enhanced multi-hop reasoning algorithm is adopted, multi-hop reasoning is carried out on the compensation updating map, symbol reasoning, vector reasoning and context reasoning are carried out in parallel in each hop, and an entity relation reasoning result is obtained through dynamic weight fusion; and superposing an entity relationship reasoning result to the compensation updating graph through a cloud collaborative node, adding an entity edge and automatically maintaining a version history log, and obtaining a reasoning fusion version knowledge graph. Through a multi-hop reasoning algorithm enhanced by a large language model, the deep semantic mining capability and reasoning accuracy of a cross-version entity relationship are effectively improved.

The invention provides an automatic focus identification and grading method for a medical image, and the method comprises the steps: carrying out the standardization of an obtained multi-modal original image based on anatomical constraint, and obtaining a standardized image; generating semantic enhancement features through a cross-modal feature compensation network based on the standardized image and associated radiological text description; performing dynamic feature adaptation processing on the semantic enhancement feature to generate a modal adaptive feature; performing context reasoning through a multi-scale feature interaction algorithm based on the modal adaptive features to generate context reasoning features; and lesion identification decoding processing is carried out on the context inference feature map, a lesion segmentation mask is generated, and the lesion segmentation mask is used for extracting lesion area feature parameters to carry out lesion classification. By adopting the method, the adaptability to the missing mode can be enhanced, and the focus identification and grading precision can be improved.

The invention relates to the technical field of artificial intelligence and big language model reasoning optimization, and discloses a big language model-oriented adaptive KV cache compression method and system, and the method comprises the steps: constructing a lexical element importance measurement mechanism; analyzing an attention head distribution structure in large language model reasoning, and constructing a plurality of pruning strategies; based on a lexical element importance measurement mechanism and the attention head distribution structure, designing a self-adaptive key value cache compression hybrid strategy set based on a pruning strategy; constructing a static self-adaptive key value cache compression method, and automatically distributing a key value cache compression strategy in a pre-filling stage of large language model reasoning; and in a decoding stage, performing adaptive compression on the key value cache based on the allocated key value cache compression strategy. According to the method, the key value cache can be efficiently compressed on the premise of not depending on explicit attention score calculation, a system-level reasoning optimization framework is compatible, the generation performance is kept, meanwhile, the video memory consumption is remarkably reduced, and the long context reasoning capability is enhanced.

The invention discloses an edge end multi-modal perception and decision cooperation method and system, and relates to the edge calculation and artificial intelligence cross technology field, and the system comprises a multi-modal perception unit used for collecting heterogeneous perception data of an environment; the data processing unit is in communication connection with the multi-mode sensing unit; wherein the data processing unit comprises a dynamic cognitive kernel, and the dynamic cognitive kernel sequentially comprises a modal credibility evaluation layer, a context reasoning layer and a decision verification layer; and the modal credibility evaluation layer is configured to receive the heterogeneous sensing data. According to the edge-end multi-modal sensing and decision-making collaboration method and system, by introducing a dynamic cognitive kernel, real-time evaluation and dynamic weight adjustment of the reliability of multi-modal sensing data are achieved, the defect that the collaboration efficiency of a traditional fixed rule system is reduced when the environment changes is effectively overcome, and the collaborative performance of the system is improved. And the adaptability and decision accuracy of the system in a complex scene are improved.

The invention discloses an adaptive and interpretable traffic situation prediction method and system based on a large language model, and the method comprises the following steps: obtaining historical traffic data related to a prediction task, inputting a trained time sequence prediction basic model, and generating candidate traffic state tracks of a plurality of prediction periods based on the prediction task; acquiring context information related to a prediction time period, and evaluating the candidate traffic state trajectory based on a configured large language reasoning model to obtain an optimal trajectory conforming to the context information of the prediction time period; and based on the optimal track and the context information, generating a text report by utilizing the configured large language interpretation model. Probability prediction is carried out through the time sequence prediction basic model, context reasoning and text generation are carried out through the large language model, the accuracy, adaptability and interpretability of traffic prediction under abnormal events are remarkably improved, and powerful support is provided for intelligent traffic.

This specification discloses systems, methods, and techniques for classifying a grade of cancer represented in a tumor tissue sample. The techniques include obtaining a whole-slide image (WSI) of the tumor tissue sample and partitioning the WSI into a collection of patches. For each patch, (i) a contextualized feature representation of the patch is generated based on intrinsic features of the patch, local dependencies between the patch and a subset of local patches of the WSI, and non-local dependencies between the patch and a subset of non-local patches of the WSI; and (ii) an attention weight is determined for the patch. A WSI-level cancer grade for the tumor tissue sample is predicted based on the contextualized feature representations and the attention weights for the plurality of patches.

The embodiment of the invention provides a knowledge graph fusion method and system based on a large model, and the method comprises the steps: carrying out the data standardization, entity feature enhancement and relation semantic annotation of heterogeneous knowledge graphs from different sources; through semantic similarity calculation, context reasoning and alignment confidence evaluation, a multi-stage and multi-mode entity alignment mechanism is constructed in combination with a large language model, and cross-source entity matching is performed on entities in heterogeneous knowledge maps of different sources; based on conflict detection, dynamic weight distribution and a conflict resolution strategy, generating a data fusion result, and unifying multi-source data; and generating a high-quality unified knowledge graph through missing relationship prediction, logic consistency verification and an incremental updating mechanism. According to the knowledge graph fusion method and device, full-process automation, precision and dynamics of knowledge graph fusion are realized, the problems of high manual dependence, weak semantic processing capability, poor cross-domain adaptability and the like in the prior art are effectively solved, and the efficiency and quality of knowledge graph fusion are remarkably improved.

The invention provides an information query method and device based on a large language model. The method comprises the steps that query information input by a user is received; determining a target context fragment related to the query information based on the semantic similarity between historical context fragments associated with the user in the large language model stored in a database and the query information and a predetermined comprehensive importance score of each historical context fragment; wherein the comprehensive importance score of each historical context fragment is determined on the basis of a clustering result and the generation time of each historical context fragment after the historical context fragments are clustered; and fusing the query information and the target context fragment, inputting the fused information into the large language model, and determining a query result corresponding to the query information. Through the method, the context reasoning capability of the large language model can be improved, and the query result can be more accurately generated for the user.

The invention relates to the technical field of artificial intelligence and intelligent auditing, and discloses an intelligent auditing method based on a multi-modal large model, and the method comprises the following steps: S1, carrying out the image preprocessing; s2, performing depth feature extraction based on a multi-modal large model; s3, dynamically adjusting OCR identification parameters; s4, performing context analysis on an OCR recognition result through a semantic understanding module; s5, expanding the training data set by adopting a data enhancement technology; s6, parallel computing and hardware acceleration technologies are utilized; and S7, converting an OCR recognition result into a structured data format. According to the method and the device, the technical effects of unified recognition and structural understanding of multi-modal document information are achieved by adopting the technical scheme based on image-text fusion modeling and hierarchical semantic analysis, and compared with the technical scheme of semantic dislocation and context disjunction caused by image and text processing separation in the prior art, the technical scheme has the advantages that the method and the device are easy to implement; the problems of low heterogeneous modal information fusion degree and weak context reasoning capability are solved.

The embodiment of the invention provides an enterprise information architecture (4A)-based AI Agent planning enhancement system, a control device and equipment, and the system comprises a Prompt construction module which is used for analyzing an unstructured demand to obtain a structured demand, and cooperating with other modules to generate a standardized Prompt comprising a business process, a data entity and a tool strategy; the knowledge matching module is used for matching related knowledge based on a knowledge graph employment number fusion knowledge base and sorting the related knowledge into a knowledge list containing business processes, practical experience, a data list and a tool list; and the context reasoning module is used for generating a reasoning result based on the historical data and the knowledge list and feeding back the reasoning result to the Prompt construction module so as to assist in generating the normalized Prompt. According to the system, the integration capability of the AI Agent on business logic and data resources in the task planning process is remarkably improved, so that the AI Agent can more accurately understand enterprise-level business scenes, and standardized and intelligent planning support is provided for practical application of the AI technology in enterprise digital transformation.

The invention discloses a multi-source heterogeneous information extraction and structured processing method based on natural gas business data, and relates to the technical field of artificial intelligence application in the energy industry, and the method comprises the steps: analyzing a multi-modal document: carrying out the content analysis of natural gas sales documents in various formats, extracting key information, and obtaining the analyzed original data; data preprocessing: cleaning, recombining and standardizing the data to obtain preprocessed data; the mixed information extraction comprises key business index extraction, field rule base establishment, mixed extraction model establishment and context reasoning, and missing items in data are complemented by analyzing overall information and local content of a document, so that the integrity and accuracy of the information are improved; performing intelligent post-processing and constructing a relational database; according to the processing method, the natural gas service data can be accurately and efficiently extracted from the documents in various formats, and a basis is provided for subsequent data analysis and decision support.

The invention relates to a game process real-time simulation system for team sports tactics. The game process real-time simulation system comprises the following steps: defining action purposes of tactical intention space, position + orientation + including pulling space and defending clamping; a team dynamic aggregation mechanism is designed, and an intention density distribution diagram is formed by clustering changes of relative layouts of players in the space; extracting tactical motivation change nodes by using a high-order state function comprising a space tension field and a collaborative index; modeling the tactics into a multi-layer semantic graph, including a role graph, a spatial intention graph and a time sequence transformation graph; introducing a graphical inference engine for judging whether the current tactical structure accords with preset logic including whether joint defense is torn and whether space is successfully occupied; performing context reasoning judgment based on situation driving conditions including scores and remaining time, and outputting whether the current tactics are effective or not; wherein the output content comprises a tactical execution score and a decision chain graph; the tactics of the two parties are expressed as a set of intention fields and resource control points.

The invention provides a gastroscope image recognition and auxiliary diagnosis system and method based on deep learning and large language model fusion, and the system comprises a data collection and processing module which is used for obtaining and preprocessing various types of gastroscope images; the gastroscope picture recognition and classification module is used for inputting the preprocessed gastroscope pictures into a GastroHAF model for training and outputting recognition and classification results of the gastroscope pictures, and the GastroHAF model comprises a weak supervision anatomical mark positioning sub-module, a space co-occurrence context reasoning sub-module and an anatomical atlas guided hierarchical classification sub-module; and the auxiliary diagnosis module is used for outputting stomach diagnosis results and examination suggestions based on the fine-tuned Chinese medical big language model. The GastroHAF model provided by the invention can quickly, effectively and accurately judge the position of the gastroscope in combination with prior anatomical knowledge, and can effectively assist doctors in stomach examination in combination with a large language model.

The invention relates to the technical field of data fusion, in particular to a water conservancy teaching scene dynamic simulation method and system based on multi-modal data fusion, and the method comprises the steps: firstly, carrying out the cross-modal feature extraction and space-time alignment of multi-source heterogeneous data, generating fusion features in a unified feature space, and solving the problem of data feature inconsistency; secondly, semantic enhancement and context reasoning are carried out on the fusion features based on a water conservancy project knowledge graph, and a physical simulation algorithm and a generative model are combined to drive and generate a physically accurate dynamic simulation scene; and finally, mapping a user operation instruction to a unified feature space in real time in teaching deduction to generate an operation feature vector, and dynamically generating a corresponding virtual intervention result through deviation analysis, thereby effectively solving the problem of interactive response delay, and realizing unification of authenticity and interactivity of teaching simulation.

The invention discloses a medical examination report processing method and device, electronic equipment and a storage medium, and the method comprises the steps: processing a medical examination report, and determining report text data corresponding to the medical examination report; processing the report text data based on a big language model, and obtaining structured data which is output by the big language model and corresponds to the medical examination report; and converting the structured data into standardized data based on a standard document format, and mapping the standardized data to a medical coding system. Based on the technical scheme, a prompt project and context reasoning mechanism is utilized to guide a large language model to automatically identify the medical entity and output the structured representation, and the structured data is mapped to a medical coding system, so that seamless joint with various medical information is realized; and the real-time adaptive analysis and structuring requirements of the multi-source medical examination report can be met.

The invention relates to a large model database query generation method, system and device based on semantic matching prompt and a medium, and the method comprises the following steps: constructing a database query-oriented natural language semantic representation model, and generating a semantic vector by taking a natural language query text input into a database by a user as input; based on a semantic vector similarity matching mechanism, retrieving a historical query text most similar to the semantic vector from a pre-stored training set; extracting a database query statement corresponding to the historical query text as a semantic guidance prompt, and forming a structured prompt word instruction with a natural language query text input into a database by a user; and inputting the prompt word instruction into a pre-trained large language model, calling the context reasoning capability of the large language model, and generating a target database query statement consistent with the intention of the user, thereby completing the conversion from the natural language to the SQL, and obtaining a user query result. The method can be widely applied to the crossing field of artificial intelligence and database technologies.

The present disclosure relates to a method and system for enhancing inference in large language models (LLMs) over long input sequences. A segmented inference strategy may be employed, wherein the long context can be divided and sequentially processed through a key-value (KV) cache of the LLM. At each step, the model may generate auxiliary outputs (or margins), which may include extractive summaries or intermediate signals based on the segment's relevance to an instruction. These margins may then be classified and selectively retained to guide final inference on the instruction. The retained margins may be prepended to the instruction to facilitate improved generation without modifying the model's internal weights. The disclosed approach provides efficient localization of relevant content, improves comprehension of extended contexts, and reduces computational overhead. Moreover, the disclosed technique is particularly effective for retrieval-based NLP tasks and supports long-context reasoning in LLMs while enhancing inference efficiency and user experience.