6 results about "Context manager" patented technology

The application discloses a kind of text context dynamic reconstruction methods, comprising: to original document is step length overlap segmentation processing, generate semantic fragment and record physical offset index information;Build document context manager storage original document complete content;Perform vector retrieval and obtain the physical offset index of matching fragment;Real-time analysis and determine the physical position boundary of fragment in original document;From original document, real-time extraction predecessor and successor text block;Splice form complete logical context window.The application establishes the accurate position anchor point of semantic fragment and original document by physical offset index, realizes zero-cost backtracking original;Adopt retrieval time dynamic reconstruction, support according to the context expansion range of adaptive adjustment of retrieval confidence;Provide three-level context mode to meet different retrieval scene demand;Intelligent deduplication eliminates redundant content.Compared with traditional RAG system, retrieval accuracy is improved by about 18%, context completeness is improved by about 31%, and Token consumption is reduced by about 30%.

A processing core for executing threads is provided. The processing core comprises: one or more functional units; a shared register file comprising a plurality of registers, each register of the shared register file being configured to store data for executing any thread; and a multi-thread elaboration unit, MTEU, the MTEU comprising: a control module configured to: receive an indication of a plurality of active threads for a given scheduling window; and receive instructions for each of the plurality of active threads; a context manager module configured to load, for each of the plurality of active threads, data required for executing a next one or more of the instructions into a respective one or more registers of the shared register file; and an instruction scheduler module configured to schedule the instructions of the active threads for execution based on availability of the one or more functional units.

The application discloses a kind of text context dynamic reconstruction methods, comprising: to original document is step length overlap segmentation processing, generate semantic fragment and record physical offset index information;Build document context manager storage original document complete content;Perform vector retrieval and obtain the physical offset index of matching fragment;Real-time analysis and determine the physical position boundary of fragment in original document;From original document, real-time extraction predecessor and successor text block;Splice form complete logical context window.The application establishes the accurate position anchor point of semantic fragment and original document by physical offset index, realizes zero-cost backtracking original;Adopt retrieval time dynamic reconstruction, support according to the context expansion range of adaptive adjustment of retrieval confidence;Provide three-level context mode to meet different retrieval scene demand;Intelligent deduplication eliminates redundant content.Compared with traditional RAG system, retrieval accuracy is improved by about 18%, context completeness is improved by about 31%, and Token consumption is reduced by about 30%.

This invention discloses a method and system for small-sample intent assessment and situational awareness based on a large language model, belonging to the field of natural language processing. It mainly includes: retrieving the most relevant few-sample examples and known intent descriptions for queries using a dynamic context manager; concatenating prompts and invoking a large language model for inference, outputting intent recognition results and discovering new intents; employing a joint training strategy of supervised contrastive learning and adversarial enhancement to refine the feature space on pseudo-labeled data; and calculating anomaly scores for samples based on Mahalanobis distance to identify out-of-distribution new intents (abnormal intents). This invention, through a two-stage hybrid framework of "generative guidance and discriminative refinement," effectively solves the technical challenges caused by data scarcity and dynamic evolution of intent in open environments, achieving highly robust, low-cost, and sustainably evolving intent discovery and recognition. It can be widely applied in small-sample fields such as military activities, cybersecurity, financial risk control, and public opinion analysis.

The application provides a type inference method and device of a programming language, the method comprising: in the development process of program code, obtaining a node to be inferred in an abstract syntax tree, the node to be inferred indicating a field to be inferred in the program code, the field to be inferred comprising one of the following: a variable, a class, a function or a parameter; obtaining a constraint related to the node to be inferred, the constraint being used to indicate a relationship involved by the node to be inferred in the abstract syntax tree; determining a type of the node to be inferred according to the constraint, the type of the node to be inferred being one of the following: a numeric type, a sequence type, a mapping type, a set type, an iterator type, a context manager type, a module type, a metatype, a null type or a callable type. The scheme of the application is beneficial to reducing the memory occupation of type inference.

The application provides a vectorization performance optimization method and system for large-scale log data, and relates to the technical field of log data vectorization, and the method comprises the following steps: determining a target embedding model assigned to a global variable; determining a target event loop in a cache, and submitting an asynchronous initialization function to the target event loop for execution to complete asynchronous network connection establishment of a vector database; when accessing a unified storage context manager attribute, triggering a lazy loading process, creating a corresponding target storage object according to a current scene type, a target embedding model and a target event loop; determining a batch processing mode according to a total number of document lists of log data to be processed; performing splitting and vectorization processing on document data by using a pre-constructed ingestion pipeline object according to the batch processing mode; and storing a processing result based on the target storage object.