146 results about "Long-term memory" patented technology

The invention discloses a universal agent evaluation method, device and equipment, a medium and a product, and belongs to the technical field of agent evaluation, and the method comprises the steps: obtaining basic parameters of an agent; based on the basic parameters, the agents are evaluated in five dimensions of long-term memory accuracy, task arrangement rationality, human setting consistency, tool calling effectiveness and content compliance, so that evaluation of the agents is achieved; wherein in the process of evaluating the intelligent agent in any one of the five dimensions, a corresponding evaluation set is constructed based on the basic parameters, the large language model is adopted to score the output of the intelligent agent on the evaluation set, and the score of the intelligent agent in the dimension is obtained. According to the method, the intelligent agent can be evaluated universally and accurately.

The invention relates to a man-machine interaction method and device, computer equipment, a storage medium and a program product. The method comprises the following steps: acquiring interaction content input by a user; scene recognition is carried out on the interaction content through a large language model, and a scene label of the interaction is determined; determining a target long-term memory partition corresponding to the scene tag from a plurality of long-term memory partitions preset for the user; inputting target long-term memory content in the target long-term memory subarea and short-term memory content stored in a short-term memory area preset for the user into an intelligent agent; the intelligent agent is used for generating reply content for the interaction content according to the target long-term memory content and the short-term memory content. By adopting the method, effective association and collaborative calling of different scene memories can be realized, and the dynamic requirement of a user on coherent interaction in multi-scene switching is met.

The invention belongs to the field of artificial intelligence and robot navigation, and discloses a visual language navigation method and system based on a cross-task incremental semantic memory graph, and the method comprises the steps that an intelligent agent executes a zero-sample visual language navigation task in a continuous environment; performing cross-modal alignment on the natural language instruction and environment observation based on a multi-modal large language model, selecting candidate waypoints and updating task progress; a semantic memory graph is constructed and dynamically updated, wherein the semantic memory graph is used for structured storage and cross-task multiplexing of scene semantic information and a spatial topological relation sensed by an intelligent agent in historical tasks; and performing global path planning and local dynamic fine tuning based on the semantic memory graph. According to the method, the problem of task-by-task forgetting in a traditional method is solved, environment understanding and task reasoning capabilities are improved by constructing structured long-term memory, and navigation precision and robustness are optimized through a global-local collaborative strategy.

The invention discloses a VLA model autonomous generalization method, system and device and a medium, and belongs to the technical field of robot control. The method comprises the steps of obtaining visual observation data and natural language task description of a current environment of a robot, generating an action sequence and controlling the robot to execute tasks; track data in the task execution process are collected, and token-level dense rewards and structured reflection information are output; storing the trajectory data, the token-level dense reward and the structured reflection information into a double-memory block; extracting task priori from the long-term memory by using an experience-driven generator, and generating trajectory data containing a sub-target chain; updating the VLA strategy model by using a reinforcement learning algorithm based on the token-level dense reward; and updating the VLA strategy model by using a supervised fine tuning algorithm based on the high-quality track data set. According to the method, the technical problem of autonomous generalization of the VLA model is solved.

The invention specifically discloses a brain-like reinforcement learning method and system based on hierarchical experience playback, and relates to the technical field of reinforcement learning and brain-like computing. The method comprises the steps that S1, observation data are collected and preprocessed; s2, initializing an experience buffer pool, an actor network, a commentator network and a corresponding target network, and performing parameter initialization; s3, exploring noise is initialized, actions are selected from the actor network according to the current state and executed, and obtained experience samples are stored in an experience buffer pool; s4, obtaining a new sample from the experience buffer pool, and updating the short-term memory pool; s5, determining whether partial experience in the short-term memory experience pool is transferred to the long-term memory experience pool or not by using an attention discrimination module; and S6, updating parameters of the actor network, the reviewer network and the corresponding target network. By adopting the method, the experience utilization rate of the intelligent agent is improved, the reinforcement learning performance is improved, and the method has wide application potential in multiple fields.

The embodiment of the invention provides a context information management method and device based on hierarchical memory, equipment and a medium, and the method comprises the steps: analyzing a task request to generate a sub-task node sequence containing a target description and acceptance function, and arranging a plurality of agents to cooperatively execute a task, and after the audit is passed, a complete execution context is stored in a long-term memory, and a structured abstract is generated and stored in a short-term memory, so that the problem that the context management mechanism of the existing multi-agent system is single is effectively solved, stable maintenance of cross-session cognitive continuity is realized, and the efficiency of the multi-agent system is improved. The method improves the execution reliability of a complex long-process task, optimizes the information storage and retrieval efficiency through hierarchical memory, reduces the risk of semantic deviation accumulation and amplification, clarifies the task target and acceptance standard of each stage, enhances the multi-agent cooperation consistency, and improves the efficiency of task execution. And the execution efficiency and the task completion quality of the intelligent agent system in a complex application scene are comprehensively improved.

The embodiment of the invention provides a context construction method and device for an agent and a storage medium, and the method comprises the steps: obtaining the current input information of the agent in one interaction round, and carrying out the analysis to generate a current semantic representation; determining a semantic association degree between the current semantic representation and the context of the current dialogue task, and calculating a long-term value score of the current semantic representation; under the condition that the semantic association degree is greater than a first preset threshold value, storing the current semantic representation into a short-term memory library; under the condition that the long-term value score is greater than a second preset threshold value, storing the current semantic representation into a long-term memory library; when the intelligent agent needs to generate a response for the current input information, searching target memory content related to the semantics of the current input information from the short-term memory library and the long-term memory library; and combining the target memory content with the current input information to form prompt information, and inputting the prompt information into a large language model to generate a response for the current input information.

The invention discloses a hierarchical visual language navigation memory enhancement system in a cross-floor scene, relates to the technical field of intelligent navigation, and adopts the technical scheme that the hierarchical visual language navigation memory enhancement system comprises three core parts, namely a hierarchical visual semantic model construction module, a memory enhancement algorithm module and a navigation system integration module. The hierarchical visual semantic model construction module is used for constructing a basic visual feature layer, a floor semantic layer and a cross-floor semantic association layer; the memory enhancement algorithm module comprises a short-term memory module, a long-term memory module and a memory fusion and update strategy module; and the navigation system integration module is used for integrating the hierarchical visual semantic model and a memory enhancement algorithm into the system. Three-dimensional space characterization misalignment and path planning error accumulation can be effectively avoided, the accuracy, environment adaptability and decision-making efficiency of cross-floor navigation of the intelligent agent are remarkably improved, and the method has important technical innovation value and application prospects.

The invention discloses a dialogue data storage method based on a big language model and a related device, and relates to the field of data storage, and the method comprises the steps: calling a memory encoder to obtain dialogue data of a user and an agent, employing the big language model to carry out key information extraction operation on the dialogue data, and obtaining key information, carrying out integrity verification and rationality verification on the key information; if the verification is passed, carrying out importance evaluation on the key information to obtain an importance evaluation value; and respectively storing the key information in a long-term memory database, a short-term memory database or a working memory database according to the importance evaluation value. The purpose of reliable storage of the dialogue data is achieved based on the large language model.

The invention discloses a multi-agent-based college entrance examination consultation method and system, and belongs to the technical field of large language model application. The problems that in the prior art, a traditional college entrance examination consultation system based on a large language model is limited in complex problem processing capacity, insufficient in personalized service capacity and lack of expert knowledge are solved. The method comprises the following steps: S1, constructing a local knowledge base containing objective data and expert knowledge extracted through a specific process; s2, designing a knowledge retrieval tool for accessing a network and a local knowledge base; s3, introducing a long-term memory mechanism to store, apply and update personalized information of the user in a cross-session manner; and S4, constructing a multi-agent collaborative architecture, analyzing user intentions according to user personalized information, scheduling expert agents, and integrating results to realize college entrance examination consultation. According to the method, the accuracy and comprehensiveness of answering are effectively improved, continuous personalized services are provided, and the method can be applied to online college entrance examination consultation.

According to the intelligent auxiliary learning method and system based on the model context protocol and the cognitive state modeling, decoupling of a model end and a tool end is achieved through the model context protocol, an error type-tool dependency topological graph is constructed through the cognitive state modeling, tool dynamic screening and pruning based on deterministic rule constraints are achieved, and the method and the system have the advantages that the method and the system are easy to implement. The Token consumption is reduced, and the accuracy of model reasoning is improved at the same time. In the aspect of safety control, semantic firewall middleware is introduced into the system, output streams are monitored in real time, illegal behaviors directly giving code answers are intercepted through text semantic analysis and code abstract syntax tree comparison, and a model is forced to turn to a thought guide mode. The system also includes hierarchical context compression to maintain long term memory, adaptive difficulty knowledge retrieval based on user cognitive states, and a mechanism to utilize code sandbox to assist verification of model reasoning logic correctness. According to the method, the behaviors of the large language model can be effectively regulated and controlled, and safe, efficient and personalized intelligent auxiliary learning is realized.

Methods for developing and managing long-term memory solutions for large language models (LLMs) within a context of providing agents of the LLM as a service are disclosed. Following task-related communications between LLM agents and users of the service, information pertaining to domain knowledge, user preferences, and success or not in completing the requested task is distilled into data samples by a reflections agent of the service. The data samples are then stored into a long-term memory database that is accessible by LLM agents in the future, such that the agents can recall information of previous interactions in order to more efficiently perform new tasks for users.

The invention provides a live broadcast long video editing method and device based on memory perception collaboration, and relates to the technical field of image processing and video editing. Voice, subtitles and picture actions are analyzed in a linkage mode on a unified time grid, low-complexity characterization of a long video is achieved through an efficient perception module, and the effect of editing the long video is achieved. Cross-fragment core characters, events and semantic clues are precipitated through long-term memory of an entity center, and current perception is fed back through memory retrieval so as to enhance consistency and robustness; on the basis, a three-mode fusion wonderness score is generated, interval optimization and boundary refinement with duration budget constraint are executed, and finally, an editing decision list is output and quickly spliced into a high-quality finely-cut video. According to the invention, the problems of large quantity of long-time live broadcast materials, low efficiency of manual reviewing and editing and easy leakage of wonderful fragments are solved, and the later production efficiency and the quality of the live broadcast content are remarkably improved.

The invention discloses a retrieval enhancement generation method and system based on long-term memory and multi-source knowledge iterative fusion. The method comprises the following steps: firstly, after user inquiry is received, triggering long-term memory database retrieval, external knowledge base retrieval and large language model internal knowledge generation in parallel to construct a multi-source candidate knowledge base, and improving result correlation by adopting a self-adaptive retrieval method based on a dynamic top-k value; secondly, inputting user query and multi-source knowledge into the large language model for iterative fusion, and generating a final answer through conflict detection, fusion processing and confidence evaluation; and finally updating the long-term memory database. The method can improve the accuracy and stability of the generated answers in a multi-round interaction and complex task scene, reduces the risk of answer one-sidedness, inference incompleteness or fact deviation, enhances the multi-source knowledge utilization ability and long-term learning ability of the system, and improves the user experience. Therefore, the comprehensive performance and reliability of the system in applications such as intelligent question answering, information retrieval assistance and text generation are obviously expanded.

The invention provides an agent memory management method and device based on a decoder memory component, and the method comprises the steps: obtaining user input data; inputting the user input data into a pre-constructed large language model to generate a user response; asynchronous monitoring is carried out on the user response, and multi-source information fed back by the user is extracted; generating a training queue according to the multi-source information; and performing incremental training on the training queue and then performing knowledge solidification to realize intelligent agent memory management. According to the method and the device, the auxiliary small model is introduced, and the long-term memory is stripped from the context window of the large language model, so that flexible management and persistent storage of the long-term memory are realized, new knowledge can be efficiently captured, and loss of time sequence structure information is avoided.

The application relates to the technical fields of artificial intelligence and man-machine natural language dialogue, and provides a general memory management method and system based on a language model, which extracts memory information needing to be memorized from dialogue content with the language model, generates temporary memory, integrates the temporary memory to obtain persistent memory about a dialogue user, and is integrated into a dialogue process of the dialogue user and the language model to form memory of the dialogue user, so that the dialogue of the dialogue system is controlled, memory type classification management is performed according to the importance of the memory information in the persistent memory, so that the persistent memory contains short-term memory and long-term memory, the importance of the memory information in the persistent memory is updated and managed, and the short-term memory and the long-term memory are forgotten and upgraded, so that flexible and efficient storage and calling of user memory independent of the model type are realized, and the humanized, continuous and personalized dialogue experience in the man-machine dialogue process is improved.

The invention discloses a dialogue memory management method, system and device and a storage medium. In the scheme, newly added dialogue data generated by dialogue interaction is received and cached to a queue to be updated; and asynchronously extracting memory information of the data in the queue in batches according to a preset period to obtain candidate memory information. And updating the long-term memory bank based on the candidate memory information. And when a new dialogue request is responded, obtaining a request context from the context window, and retrieving related memory information from the updated long-term memory library according to the request context. According to the method, memory updating and real-time response are decoupled through a cache and asynchronous batch processing mechanism, the response delay is remarkably reduced while the memory accuracy and consistency are guaranteed, and the resource utilization efficiency and the system stability are improved.

The invention belongs to the technical field of software development, and particularly provides a research and development-oriented long and short-term memory framework construction method and system, which adopts a layered memory architecture to construct four core modules including a short-term memory compressor, a medium-term memory aggregator, a long-term memory graph and a cross-layer memory router. The system takes multi-source input such as research and development dialogues, code snippets and project documents as a starting point, extracts research and development elements through semantic analysis and entity recognition technologies, compresses lengthy dialogues into structured short-term memory by utilizing an attention distillation mechanism, upgrades high-frequency short-term memory into medium-term knowledge fragments based on a time sequence attenuation algorithm, and improves the research and development efficiency. And constructing a long-term knowledge graph containing developer portraits, project dependence and normative standards by adopting a graph convolutional network. Context understanding accuracy, multi-round dialogue continuity and personalized service quality of a large model in a research and development scene are remarkably improved, and the method is suitable for mainstream research and development tool scenes such as IDE plug-ins, code review and architecture design.

The invention discloses a large language model agent collaborative decision-making method for a complex dynamic game scene, and the method comprises the steps: achieving the environment perception, experience accumulation and knowledge calling functions, and supporting the cognitive modeling and strategy generation of an agent; based on cognitive information, intelligent agent role division and labor division are achieved through an action characterization device and a role selector, and the intelligent agents are guided to perform their own functions in the collaboration process. Cognitive information and role information are input into a large language model, hierarchical analysis is performed on a game situation depending on natural language understanding and thinking chain reasoning ability, key game nodes are identified, opponent strategies are predicted, and foresight collaborative decisions are generated in combination with teammate intentions. And through a semantic matching and action mapping mechanism, converting a natural language decision generated by reasoning into a structured executable instruction, and performing rationality verification. The intelligent agent executes actions and interacts with the environment, the system updates short-term memory based on feedback and periodically integrates the short-term memory into long-term memory, and a closed-loop process of'cognition-role allocation-reasoning-execution-updating 'is formed. According to the method disclosed by the invention, a distributed collaborative decision-making architecture based on a large language model is constructed, so that the intelligent agent has stronger autonomous perception, reasoning and collaboration capabilities, and the collaboration efficiency, game adaptability and strategy generalization capabilities of the intelligent agent in a complex dynamic game environment are remarkably improved.

The invention discloses an interactive response method, device and equipment for multi-modal sensing data and a storage medium, and relates to the technical field of intelligent interaction.The interactive response method for the multi-modal sensing data comprises the steps that the multi-modal sensing data is obtained, and a user mental evaluation result is constructed based on the multi-modal sensing data; identifying user identity information according to the multi-mode sensing data, and determining short-term memory information and long-term memory information according to the user identity information; when a demand instruction of a user is received, performing retrieval in the short-term memory information and the long-term memory information according to the demand instruction to obtain a retrieval result; and generating interaction response content according to the retrieval result and the user mental evaluation result, and performing interaction. Through multi-modal perception and dynamic memory management, long-term continuous personalized interaction is realized, and the user interaction experience is improved.

The application provides a memory system of an AI intelligent agent and an updating method and a retrieval method thereof, the memory system comprising a time limit query cache module, a short-term memory module, an intermediate-term memory module and a long-term memory module; the time limit query cache module is used for storing data in a short time, and the short-term memory module, the intermediate-term memory module and the long-term memory module store data with different capacities and different representation modes respectively. The application extracts historical dialogue information, stores the information in short-term memory, intermediate-term memory and long-term memory, recalls relevant memory content according to inquiry content when a user inquires, enhances prompt word information, and improves the complex task execution effect of the intelligent agent.

The invention provides an enhancement method for a question answering system during testing based on reinforcement learning, and belongs to the technical field of natural language processing. The method aims at solving the technical problems that an existing large language model (LLM) faces knowledge blind areas, reasoning chain breakage and context length limitation in a question and answer (QA) task, an existing fine adjustment method is high in calculation cost and damaged in generalization ability, and a prompt strategy seriously depends on a limited context window and lacks long-term memory. The method comprises the following steps: firstly, collecting reflection experience through multiple attempts and a failure reflection mechanism, and constructing an experience library; thirdly, performing text embedding, clustering and semantic abstracting on experiences in the experience library, and constructing an external memory library; secondly, a memory selection process is formalized into a Markov decision process (MDP), reinforcement learning (such as a PPO algorithm) is used for training a strategy agent, and the agent is optimized with a current task as a state, with memory item selection as an action and with LLM answer correctness as a reward; finally, in a test time (inference) phase, the agent dynamically selects the most helpful memory entry according to the new task and integrates it into the hint of the LLM to generate a final answer. According to the method, under the condition that LLM internal parameters do not need to be updated, the question and answer accuracy is dynamically improved, the calculation overhead is reduced, and the limitation of a context window is effectively overcome.

A system and method for implementing persistent cognitive computation through geometric representation of thought in a dynamic latent manifold. The system encodes inputs into a curved space characterized by time-evolving metric tensors, compression pressure fields derived from Ricci curvature, and goal potential fields that shape attention flow. Cognition occurs through geodesic traversal of this manifold, with attention following paths that minimize cognitive action while balancing semantic density and goal relevance. A Cognitive Dynamics Engine maintains manifold geometry, computing optimal trajectories and managing thought bundle operations including consolidation, expansion, and higher-order abstraction. During idle periods, autonomous dreaming processes reorganize the manifold through perturbation, recombination, and topological surgery. This architecture enables persistent memory through geometric encoding, where frequently accessed concepts develop high-curvature regions and cognitive shortcuts emerge from usage patterns, transforming artificial intelligence from stateless computation to structured motion through shaped memory space.

This invention relates to a method and system for mining potential user needs based on hierarchical temporal memory enhancement. First, a dynamic time window hierarchical slicing mechanism is used to decouple the cleaned multimodal user behavior sequence into long-term historical sequences and short-term real-time sequences. A semantically approximate nearest neighbor graph is constructed from the long-term historical sequence, and adaptive community discovery is performed. This is combined with a large language model to generate a high-semantic-density long-term memory abstract representation. Then, a hybrid prompt template integrating system instructions, long-term memory representation, and short-term sequences is constructed to drive the large language model to complete a three-stage thought chain reasoning process: focused denoising, scene instantiation, and function completion, generating a structured reasoning data package. A dual-tower semantic retrieval and scene-constrained filtering architecture is employed to achieve accurate mapping between the reasoning results and the product database, outputting a product list with recommendation reasons. This invention endows the recommendation system with deep logical reasoning and cross-domain association capabilities, improving the accuracy and interpretability of e-commerce recommendations.

The invention discloses a multi-modal visual language navigation method based on a dynamic environment knowledge graph and related equipment, and can be applied to the technical field of artificial intelligence. After the real-time environment image of the area to which the target robot belongs is captured through the single camera, feature extraction is carried out on the real-time environment image to obtain the image features, the to-be-processed category corresponding to the target object is coded to obtain the semantic code, and after the spatial features of the target object are extracted, the target robot is obtained. Constructing a dynamic environment knowledge graph according to the confidence, semantic coding and spatial features of the target object in combination with an external knowledge base, extracting high-dimensional graph features of the dynamic environment knowledge graph, and performing implicit fusion and display modeling in combination with image features and language features corresponding to the target instruction to obtain a navigation state vector containing long-term memory; therefore, the moving operation of the target robot can be controlled based on the navigation state vector, efficient and explainable navigation reasoning is further realized, and the language navigation accuracy is improved at relatively low cost.

The present application relates to the technical field of intelligent health management, and particularly relates to a multi-modal intelligent health management method and system based on continuous memory, comprising: multi-modal preprocessing of voice, text, images and sensor data input by a user to obtain multi-modal health data; adopting a hierarchical continuous memory storage strategy to obtain a long-term memory item set through three-layer memory architecture step-by-step compression processing; constructing a health memory graph network based on a memory consolidation and association integration strategy; adopting a global health portrait incremental update and memory portrait driven parameter mapping strategy to map a health portrait into a role parameter vector; adopting a memory retrieval and knowledge graph enhanced reasoning strategy to combine historical memory recall, knowledge graph reasoning and drug information management results to generate and output a health management response. The present application realizes the organic unification of continuous memory accumulation, personalized accompaniment and intelligent health reasoning for a health management scene.

The invention relates to the technical field of data processing, and discloses a session data processing method, device and equipment and a readable storage medium, the session data processing method comprises the following steps: determining a theme tag based on session data of a user; matching the topic tag with a keyword index in the long-term memory data, and screening out a candidate memory data set meeting a preset condition; sorting the candidate memory data set according to the keyword correlation and the time information to construct a short-term memory data set; historical memory content related to the current session data is retrieved in the short-term memory data set, and target session content is generated based on a retrieval result. Through a subject-driven classified storage and preloading mechanism, the memory retrieval efficiency is remarkably improved, the repeated access overhead of a long-term memory library is reduced, the response delay is effectively shortened, the context coherence and individuation ability of multiple rounds of dialogues are enhanced, and the user experience is comprehensively improved.

The invention relates to a social subject memory simulation system and method based on a large language model, and belongs to the technical field of computers. According to the method, online and offline multi-source information fusion is realized, and behavior and situation characteristics of social subjects can be comprehensively described; a large language model semantic comprehension capability is introduced, so that the intelligent level of memory retrieval and information fusion is improved; dynamic dump and index retrieval of long-term memory are supported, so that subjects can keep semantic coherence and behavior consistency in multiple rounds of interaction; the authenticity of information spreading and interaction between social subjects is enhanced, and technical support is provided for an intelligent social system, virtual human interaction and social simulation.

The invention belongs to the technical field of natural language processing and information retrieval, and relates to an intelligent question answering method based on structured semantic indexing and double-layer memory enhancement. The method comprises six steps of query preprocessing, Agent-based multi-tool dynamic routing, structured and semantic enhanced index construction, self-adaptive context assembly, double-layer memory management, and data closed loop and self-evolution, a differential segmentation strategy is adopted for texts, codes and multi-modal data, query optimization, intelligent tool routing and a mixed retrieval algorithm are combined, and the multi-modal data are subjected to self-adaptive context assembly. The technical defects of semantic rupture, low retrieval accuracy, no long-term memory and lack of self-optimization of an existing RAG system are overcome. According to the method, the code retrieval accuracy and the context utilization rate are improved, personalized long-term service and automatic operation and maintenance are achieved, and the method is suitable for scenes such as technology research and development, code library maintenance and intelligent question and answer.

The application discloses a kind of based on agent long-term memory and optimization search's streaming video compression system, including: memory graph processing module, no edge minimum maximum sampling module and edge perception weighted pruning module and time decay memory search module, the structural characteristics of the present application based on memory graph, text node in long-term memory is distinguished according to whether it is connected with entity node, and differential compression strategy is used, effectively removes redundant memory node, while maintaining the integrity of key information is significantly reduced in storage, based on time decay memory search mechanism, different time segments of memory are given dynamic weight, pay more attention to recent memory more relevant to current query in search stage, so as to alleviate the performance loss caused by memory compression, while improve retrieval efficiency.Through the above technical scheme, the application realizes efficient compression and fast retrieval of agent long-term memory under the premise of guaranteeing streaming video understanding accuracy, significantly improves the real-time performance and scalability of the system.