222 results about "Backtracking" patented technology

The invention discloses a law compliance auditing response method driven by an enterprise knowledge graph, and relates to the technical field of artificial intelligence, and the method comprises the following steps: carrying out semantic analysis on a law and regulation text, constructing a context hierarchical expression model, dividing each law and regulation into a title semantic block, a subject semantic block, a limit semantic block and a constraint semantic block, and generating a hierarchical semantic graph. According to the method, the structured analysis of the regulation provisions is realized by constructing the hierarchical semantic graph, and the regulation provisions are accurately aligned with the business entities in the enterprise knowledge graph, so that the reasonable regulation-enterprise semantic mapping relationship is constructed. And boundary reasoning, confidence scoring, context backtracking and semantic correction are combined to realize intelligent discrimination and correction of the applicability of the regulatory obligations. Models and rules are continuously updated through a feedback optimization learning mechanism, the dynamic adaptive capacity of the system and the accuracy of compliance suggestions are improved, the misjudgment risk is remarkably reduced, and the intelligence and the practical application value of the system are enhanced.

The invention discloses an enhanced large language model processing method, system and platform based on Monte Carlo tree search and a storage medium. The method comprises the following steps: creating a first model corresponding to a big language, generating and obtaining a data set corresponding to a logical reasoning process, and performing fine tuning processing on the first model according to the data set; dividing the search tree into at least two initial paths based on task requirements, allocating at least two nodes to each path for independent sub-tree expansion, generating at least two candidate outputs based on a first model in each sub-tree, and dynamically adjusting and processing a corresponding cluster width in combination with confidence data; independently simulating and processing the expansion path of each sub-tree, generating corresponding first data, backtracking, updating and processing confidence data of each node according to the first data, and generating corresponding second data at the same time; wherein the second data is the reasoning response data after enhancement optimization processing, and the corresponding system, platform and storage medium can improve the reasoning processing efficiency and accuracy.

The invention discloses a multi-mode collaborative digital human interaction method and related equipment based on a thinking chain, and the method comprises the steps: receiving language, action and text signals inputted by a user through a multi-mode perception module, and generating a multi-mode thinking chain driven by the thinking chain in combination with a historical dialogue scene; the thinking chain dynamically binds time sequence nodes of voices, actions and expressions and defines semantic association of the time sequence nodes; emotional consistency, time sequence coherence and intention matching degree of multi-modal output are verified through a real-time cooperative verification mechanism, and local backtracking and dynamic correction of a thinking chain are triggered; and finally, driving the digital person to output collaborative voice, actions and expressions according to the optimized thinking chain. According to the method, a multi-modal interaction thinking chain is generated by adopting a large language model, semantic deep collaboration and dynamic adaptability of digital human output are realized through a sequential binding and real-time verification mechanism, the problems of multi-modal splitting, intention matching deviation and response stiffness in the traditional technology are solved, and the interaction simulation degree and scene robustness are remarkably improved.

The invention belongs to the technical field of data desensitization, and provides a database sensitive data desensitization and backtracking control method and system.According to the technical scheme, text samples are constructed according to obtained database field content, sensitive entities in the text samples are recognized, and the sensitive entities are classified and marked; determining a field position and a sensitivity level of the sensitive entity; determining a desensitization strategy by combining three-dimensional information of a field type, use scene information and user role information on the basis of the determined field position and sensitive level of the sensitive entity; and responding to an operation request of a user, matching a corresponding desensitization strategy and an execution strategy based on the operation request of the user, writing a strategy execution record into an audit log, performing data use track backtracking and anomaly detection according to the audit log, and obtaining a corresponding disposal strategy according to an anomaly detection result. The data security governance capability can be remarkably improved, and the requirement of supervision on data compliance use is met.

The invention discloses a high-precision instrument assembly fault backtracking method and system, belongs to the field of precision manufacturing, and aims to solve the problems that in a traditional backtracking method, assembly data are scattered and unreliable, fault root positioning is fuzzy, and new scene adaptation depends on a large amount of data. The method comprises the following steps: collecting assembly structured data, video images and environment data in a multi-source manner, filtering out low-quality images, and distributing unique identifiers for products; fusing the multi-modal data to generate a depth feature matrix; constructing an anomaly detection model to output a risk score and a label; hashing the data and then storing the data into a product exclusive private block chain; when a fault occurs, extracting data on the chain through a unique identifier, reconstructing an assembly process by using a graph neural network, and comparing a standard positioning root; and based on the fault report incremental training model, parameters are optimized in combination with meta-reinforcement learning. According to the method, the data authenticity is guaranteed, the fault backtracking precision and efficiency are improved, a new scene is quickly adapted, the production rework rate is reduced, and the stable assembly quality is maintained.

The invention provides a resource unified management method and system based on heterogeneous computing power, and relates to the technical field of resource allocation of heterogeneous computing power. According to the method, operation instructions of different computing units are converted into bidirectional irreversible instruction sequences by constructing a heterogeneous presentation layer, and physical resources are abstracted to generate a virtual computing power unit. A multi-dimensional efficiency baseline is constructed based on parameters such as real-time load and network jitter tolerance, and a time-varying efficiency curve is dynamically generated in combination with a hardware aging rate. A user task is decomposed into sub-task chains, intelligent resource binding is realized by matching hardware sensitivity characteristics with an efficiency curve, and conflicts are solved by adopting reinforcement learning. And monitoring efficiency offset in real time during execution, performing reverse backtracking during abnormality, and generating a recombination strategy based on the alternative resource set. According to the system, efficient management and dynamic optimization of heterogeneous computing power resources are realized, and the computing efficiency and reliability are remarkably improved.

The invention discloses an automatic driving system efficient fuzzy test method based on system slicing and backtracking, and the method is based on a data-driven scene library, takes a real-time driving behavior value-at-risk as a feedback signal, achieves the rapid generation and evolution of a high-risk scene through a system state slicing and backtracking mechanism, and achieves the real-time driving behavior value-at-risk. Comprising the steps of introducing a genetic algorithm regulation and control mechanism of multi-source risk information fusion, branch path aggregation evaluation and risk guidance, and a population evolution strategy based on dynamic feedback, so that a self-driven closed-loop mechanism of a test system from generation-test-evaluation to perception-feedback-re-optimization is realized. The method can improve the recognition capability of the automatic driving system for the high-risk long-tail scene and the system-level fault exposure efficiency, and is suitable for the scenes of software testing, simulation verification, risk assessment, safety strategy optimization and the like of the automatic driving system.

The invention discloses a large language model knowledge base self-adaptive updating method combining documents and interactive feedback, and relates to the technical field of knowledge base updating, and the method comprises the steps that a real-time change detection module monitors streaming documents and interactive feedback, and captures and marks a single fact change event; the decision module calculates a shunting decision coefficient according to the heat density index and the linkage sweep index, and the event is shunted to a Micro-Edit or Incremental-Tune path; if shunting to the Micro-Edit path, the positioning module executes attention backtracking and gradient sensitivity analysis, and determines a to-be-updated weight address; the rank-one editing module implements rank-one updating and records an audit log; the verification module generates an online token or a rollback instruction through consistency comparison; the version management module manages a version, activates a new version route and cleans out expired branches. Through event granularity identification, intelligent shunting, local updating and multi-version management, the timeliness of knowledge base updating is remarkably improved, and resource consumption and maintenance cost are reduced.

The invention relates to the technical field of large language models, and discloses a zero-illusion large language model output method and system, and the method comprises the steps: sequentially carrying out the deterministic lexical analysis and standardized mapping of a received original query, and generating a standard term set; locking to-be-reasoned items through hierarchical matching to form a deterministic retrieval result set; executing five rounds of progressive verification processing on the result set to obtain an integrated result set with an LLM verification mark; and outputting a layered zero illusion result through multiple verification and auditing of structure compliance, semantic consistency and knowledge base double recheck. According to the method, probabilistic reasoning is replaced by full-link deterministic operation, so that each code can be traced back to a credible knowledge source, illusion is thoroughly eliminated, absolute reliability and complete traceability of an output result in a professional field are realized by a verifiable deterministic path, and a foundation is laid for zero-illusion application.

The invention relates to a program debugging method and device, equipment and a medium, and the method comprises the steps: responding to a program execution error event, and determining incremental data and memory change information between every two program state snapshots with adjacent time sequences in a process of executing a program from an error starting point to an error backtracking point; determining the corresponding program state as the current program state based on the program state snapshot corresponding to the error backtracking point; according to the current program state and the corresponding incremental data, it is determined that the program state corresponding to the previous program state snapshot is updated to be the current program state, and corresponding program state information is obtained; and the steps are iterated until the current program state is traced back to the program state corresponding to the program error starting point, iteration is ended, and a trace-back detail record containing program state information is visually fed back to the user. According to the method, a backtracking detail record is provided for a developer by utilizing a reverse backtracking debugging method so as to assist in quickly positioning and analyzing an error source.

The invention discloses an automatic verification excitation generation method based on a backtracking thinking tree, and the method comprises the steps: constructing a backtracking thinking tree frame based on a self-feedback mechanism based on a thinking tree for a to-be-verified processor function; under the backtracking thinking tree framework, decoupling a to-be-verified function from top to bottom through a large language model, and recursively decomposing the whole to-be-verified function into function points capable of being independently verified layer by layer through a tree structure; according to each function point, determining a normal function and a boundary of each function point, and for each function point, generating verification excitation layer by layer in a tree structure; after the verification excitation is generated, self-verification and self-backtracking are carried out layer by layer through a tree structure, and finally, the generated verification excitation is executed on the processor design to verify the correctness of the processor function. According to the method and the system, the function verification efficiency and the verification coverage rate of the processor are remarkably improved.

The invention discloses a large language model reasoning acceleration method and device based on two-stage speculative decoding and a storage medium, and the method comprises the steps: constructing and initializing a Trie tree, and inserting a historical corpus, and phrase sequences in a document library or a code library into the Trie tree one by one; in the reasoning process, longest prefix matching is carried out based on a Trie tree, and a candidate draft sequence is generated by adopting branch backtracking and recursive search; performing confidence evaluation on the candidate draft sequence, calculating a joint confidence score of the sequence through probability multiplication and a Top-K screening mechanism, and judging whether the joint confidence score reaches a confidence threshold; if the accumulated confidence of the candidate sequence reaches a threshold value, skipping a small model generation stage, and directly entering large model verification; otherwise, entering a small model draft completion stage; and the final large model takes the replaced and updated draft sequence as final output. According to the method, adaptive acceleration of the decoding process can be realized, and the long text reasoning delay of the large language model is remarkably reduced while the generation quality is ensured.

The invention provides a multilevel backtracking unmanned aerial vehicle path planning method based on a neighborhood extension algorithm, aiming at the problems that the traditional A * algorithm is low in efficiency and easy to fall into dead angles in unmanned aerial vehicle path planning. The method comprises the following steps: firstly, generating a random two-dimensional grid map simulation task environment; secondly, grids in eight neighborhoods of the current position of the unmanned aerial vehicle are scanned in real time, and local environment information is dynamically updated; and finally, a neighborhood extension planning algorithm is adopted, the unmanned aerial vehicle is allowed to perform diagonal motion, the planning flexibility and efficiency are improved, and redundant nodes and unnecessary curves are reduced. According to the method, a multi-stage backtracking strategy is designed, and a key backtracking point is preset in a path planning process, so that the unmanned aerial vehicle can quickly identify and return to a preset optimal backtracking point when entering a dead angle due to the limitation of a local environment, thereby quickly returning to a safe area and re-planning a path. According to the strategy, local optimal solution and infinite loop search are effectively avoided, and the task execution efficiency is remarkably improved in a complex environment containing a large number of dead angles.

The invention discloses a Token conversion method and system for converting multi-modal information to a unified control instruction, and the method comprises the steps: obtaining a multi-modal perception information flow and instruction reference data, carrying out the feature deconstruction of the multi-modal perception information flow, recognizing a modal identification code, carrying out the cross-domain mapping calculation based on the modal identification code and the instruction reference data, generating an offset vector, and carrying out the Token conversion of the multi-modal information to the unified control instruction. Performing tensor synthesis on the offset vectors to construct a Token coding library; performing intention flow direction analysis according to the Token coding library to generate an intention propagation domain, performing semantic potential energy optimization on the propagation domain to form an optimal intention link, and generating an adaptive Token flow; the input analysis layer, the logical reasoning layer and the output mapping layer are positioned and recognized through semantic backtracking, semantic nodes are implanted, node balance adjustment is executed, and a cooperative processing chain is formed; feature reverse extraction, reasoning complementation processing and deep reasoning processing are executed through a cooperative processing chain, a three-layer control network is formed, a control Token sequence is generated, and efficient conversion from information to instructions is achieved.

The invention discloses an interaction state backtracking and semantic reconstruction system based on a user operation log. The system comprises an operation log recording module which is used for capturing interactive operation of a user in an interface in real time and generating an event log with a timestamp; the state backtracking module is used for recovering the interactive semantic state of any time node according to the log; the semantic reconstruction module is used for allowing a user to derive a new interaction path on the basis of the backtracking state so as to realize branch evolution of a semantic structure; and the user control confirmation module is used for providing a confirmation mechanism in the state recovery and reconstruction process to ensure that final execution is triggered by a user. The technical problems that an existing system lacks historical operation tracking, and semantic rollback and reconstruction are difficult to carry out are solved.

The invention discloses a self-defined component configuration method and system based on a low-code platform, and belongs to the technical field of low-code development, and the method comprises the following steps: constructing a reusable dynamic structure unit library by deconstructing a component function into an atomized dynamic structure unit and defining metadata and context contracts of the dynamic structure unit; component functions are disassembled into template fragments, logic controllers and metadata descriptors in an atomized mode, standardized metadata and context contracts are defined, a reusable dynamic structure unit library with high cohesion and low coupling is constructed, label retrieval according to function categories is supported, a version control mechanism guarantees historical version backtracking and automatic upgrading, and the reliability of the system is improved. Resource packaging simplifies an integration process for a standardized module, so that the repeated development cost is remarkably reduced, and an efficient and extensible unit basis is provided for dynamic configuration.

The invention discloses a risk policy verification method, device, system, equipment and medium, and the method comprises the steps: obtaining a plurality of risk policies to be verified and a business event data set of a plurality of business events, and storing the plurality of risk policies and the business event data set in a sandbox environment; generating a plurality of groups of simulation strategy schemes corresponding to each business event based on the plurality of risk strategies and the at least one execution strategy in the sandbox environment; performing backtracking on each service event based on the request parameter of each service event and the multiple groups of simulation strategy schemes in the sandbox environment to generate a backtracking result; and verifying the plurality of risk policies based on the backtracking result and the response result in the sandbox environment. Through the above mode, on the basis of batch automatic backtracking verification of real business data, the risk strategy verification efficiency is effectively improved on the premise of ensuring the data isolation security.

The invention relates to the technical field of access control, in particular to an enterprise trusted data authorization and backtracking method based on a dynamic two-dimensional code, which comprises the steps of setting an enterprise access authority rule, generating a short-time effective two-dimensional code in combination with a dynamic time factor and a key, and supporting security code scanning access control. Behavior feature vector modeling and topological entropy analysis are introduced for behavior scoring, so that the exception recognition capability is improved; access logs are collected to construct a chain structure, and traceability and privacy security of the whole data access process are realized in combination with zero-knowledge proof and differential privacy protection. According to the method, the two-dimensional code containing the access permission and the key identifier is dynamically generated, so that real-time authorization control and full-link behavior tracing of enterprise data access are realized, and the security of data access is guaranteed.

The invention discloses a semantic unification and anomaly traceability analysis method and system for heterogeneous program logs, and the method comprises the steps: firstly collecting original logs in various formats, and packaging the original logs into standard objects in a unified manner; thirdly, identifying fields through a rule and a semantic model, uniformly mapping and labeling semantic tags, eliminating semantic differences, and constructing log data with uniform semantics; automatically constructing a cross-system call chain based on the request identifier and the like; secondly, extracting log features and comparing the log features with a normal behavior model to carry out anomaly detection, and generating an abnormal event; combining time, calling and dependency to construct an anomaly propagation path, and reversely backtracking and positioning an anomaly source; abnormal priority scores are calculated according to the propagation range, the influence degree and the like, and sorting and visual display are carried out; through unified semantic modeling and automatic association analysis of heterogeneous logs, accurate tracing and intelligent positioning of cross-service link anomalies are realized, the operation and maintenance efficiency is effectively improved, and the dependence on artificial experience is reduced.

The invention relates to a dialogue long context precision optimization method based on abstracts and indexes, which mainly solves the problems of information loss and storage efficiency in a long dialogue scene by dynamically generating dialogue abstracts and content indexes. The technical scheme comprises the steps of collecting and preprocessing dialogue content; generating a dynamic abstract; hierarchical storage construction; constructing an intelligent index; a content backtracking mechanism; and performing adaptive optimization.

The invention provides an intelligent course arrangement method based on multi-constraint hierarchical optimization and dynamic backtracking, and is applied to the technical field of data processing. According to the method, basic data of setups, classes, courses and the like are firstly obtained, and a multi-dimensional constraint model is established through preprocessing and is quantified into a time slot state value; based on a two-level layering strategy and dynamic teacher priority scheduling, independent classes are processed firstly, then sub-classes are processed, conflicts are cleared, and a stage course arrangement result is generated. In combination with the weekly class hours of the courses, time is distributed through an intelligent hall connection and degradation strategy, conflicts are detected in real time, and a backtracking mechanism is triggered. And verifying the conflict-free persistent data, and generating an initial course arrangement scheme. And finally, according to indexes such as course arrangement success rate and the like and application feedback, a constraint model, a scheduling algorithm and a class connection strategy are optimized, a multi-dimensional coupled intelligent course arrangement optimization scheme is formed, and the course arrangement problem under complex constraints is efficiently solved.

The invention relates to the technical field of data processing, and discloses a computing power server system layer optimization method and system oriented to a high-load scene. The method comprises the steps of collecting micro performance indexes such as the CPU instruction cycle number and the page table missing frequency, constructing a three-layer causal directed acyclic graph through Granger causal inspection, reducing a parameter search space based on bottleneck node reverse backtracking, generating an interpretable optimization decision with a causal path and contribution degree quantification, and carrying out optimization on the basis of the interpretable optimization decision. The problems that the performance bottleneck root cause cannot be accurately positioned and the optimization result lacks transparency in the prior art are solved. According to the method, bottleneck node reverse backtracking and parameter space pruning are performed based on the causal atlas, so that the problem of low optimization efficiency caused by incapability of accurately positioning a performance bottleneck root cause and blind exploration of a parameter space in the prior art is solved.

The invention provides a backtracking-based spatial coupling low-density parity check code decoding method. The backtracking-based spatial coupling low-density parity check code decoding method comprises the following steps: acquiring a code word sequence of a spatial coupling low-density parity check code to be decoded according to set parameters; performing side information initialization on the decoding window, executing belief propagation decoding in the decoding window, and performing judgment when the number of iterations reaches the set minimum number of iterations; and determining the decoding position and the window size of a decoding window in the next step according to a judgment result until all target symbols in the code word sequence are decoded. According to the invention, the error propagation phenomenon in the space coupling LDPC code sliding window decoding process can be effectively improved, and the sliding window decoding performance is improved.

The invention relates to the technical field of software configuration management, and discloses a monorepo-based multi-version control and management method, which comprises the following steps: receiving a version switching instruction containing a target global version identifier and an environment identifier; loading the physically isolated environment version configuration file from a preset path according to the environment identifier; a target configuration node corresponding to the target global version identifier is positioned in the configuration file, recursive branch information merging operation is executed based on inheritance attributes and branch attributes in the node, and a final branch configuration set containing all sub-module target states is generated; and traversing the final branch configuration set, and calling a version control tool to switch the working area state of each sub-module to the corresponding target sub-module version identifier. Through the environment isolation and increment inheritance model, the problems of version definition redundancy and environment confusion in a multi-module warehouse are solved, and accurate backtracking and atomization switching of sub-module states are achieved.

The invention belongs to the field of industrial control safety, provides a PLC (programmable logic controller) memory behavior analysis-based control system configuration logic anomaly detection method, and solves the problems of logic attack detection blind areas, incomplete detection coverage, process semantic deficiency and the like caused by hidden attacks (modifying PLC control logic through a legal programming interface). The method comprises the following steps: S1, collecting a PLC register with a timestamp and a coil state sequence; s2, identifying a key memory address, constructing an address cluster set, and supporting event completion and process mining; s3, generating an XES standard event log by adopting a clustering and symbol back-pushing method; s4, improving a Heuristics Miner algorithm, automatically identifying a starting point and an end point of a control flow, and constructing a dependency graph with an identifier; s5, converting the control logic into a formalized model, and defining a legal state transition sequence; and S6, monitoring a real-time event stream on line, mapping the real-time event stream to the model and dynamically playing back the real-time event stream, and realizing anomaly detection by combining track alignment and comparison. The method is complete in coverage, clear in semantics, capable of tracing attack paths and capable of improving the defense capability of a logic layer.

The invention discloses a metadata version control and automatic structure comparison method and system based on incremental snapshot and intelligent decision, and the method comprises the following steps: managing a metadata version through an incremental snapshot model, and enabling the incremental snapshot model to support the complete backtracking of a historical structure through storing changed incremental information and associating with a version chain; carrying out difference analysis on the metadata structure by adopting a hierarchical progressive comparison algorithm; an intelligent DDL decision engine is adopted, and an executable DDL script is generated according to the difference analysis result; in the process of generating the DDL script, metadata consanguinity information is jointly checked so as to carry out change risk assessment. Through collaborative design of all the modules, double benefits of systematic breakthrough and conventional effect enhancement are brought, and the technical effect is directly associated with collaborative design.

The invention relates to the technical field of code security, in particular to a data stream compliance detection method and device based on code semantic rule-driven LLM, and the method comprises the steps: carrying out the normalization processing of a to-be-detected source code: removing annotation, null lines, unifying branch circulation into an if-else structure, and establishing lt; code lines and line numbers gt; performing mapping; constructing a code semantic rule knowledge base; generating a code intermediate expression by using LLM; generating an execution path for the intermediate expression, constructing a symbol table, and reasoning and positioning non-compliant variables and corresponding code lines according to a data flow rule; generating a detection report; the device comprises a code standardization processing function module, a knowledge base construction function module and other function modules for achieving the method, manual intervention is not needed, C / C + + code compliance can be automatically detected, the detection precision and efficiency are remarkably improved, and non-compliance positions can be accurately traced.

The invention discloses a power grid vulnerability micropatch generation method, system and device based on call chain backtracking and a medium, and belongs to the technical field of network security, and the method comprises the following steps: based on a dynamic probe, realizing kernel layer protocol analysis, lock-free acquisition of system call and user state call chain reconstruction; key path marking is realized through LLVM-IR semantic analysis and power business feature matching; generating a security micropatch based on the metadata and realizing isolated hot loading; and the patch compatibility is ensured by adopting digital simulation verification and a layered release mechanism. The method has the advantages that a five-dimensional cooperative power grid active defense system of observation-modeling-reasoning-evaluation-arrangement is constructed, millisecond observation is achieved through eBPF collection synchronized with PTP, a multi-domain ontology atlas supports cross-domain reasoning, a GNN-RL model predicts an attack path, protection logic is verified through double-target-range simulation, and the method is high in reliability and high in reliability. The intention-driven mechanism realizes second-level response, and a complete closed loop from threat perception to adaptive protection is formed.

Methods, non-transitory machine-readable mediums, and system to provide historical positioning information are described. In an embodiment, a device detects one or more conditions in user contextual data that trigger collection of a set of historical positions for a lookback window, the device receives at least one historical position from the set of historical positions for the lookback window, the device classifies the at least one historical position based on one or more features as a candidate location for a backtrack route, and the device determines to provide the at least one historical position as part of the lookback window based on the classification.

The invention discloses a multi-agent collaborative power station index anomaly root cause step-by-step tracing method, and relates to the technical field of power system operation monitoring and intelligent fault diagnosis, and the method comprises the steps: collecting the operation index data of power station multi-level equipment, and constructing an index association topological structure based on the causal relationship between indexes; receiving a traceability request packet identified by the abnormal index, initializing an analysis agent of a corresponding level, and triggering collaborative analysis by taking the abnormal index as a traceability starting node; the analysis agent performs step-by-step traversal analysis on the upstream indexes with causal association along the index association topological structure to obtain an abnormal influence evaluation result; and aggregating the analysis results of each hierarchy, forming a causal propagation path from the abnormal indexes to the candidate root cause indexes, and outputting corresponding index identifiers, influence evaluation results and hierarchy information. According to the method, the power station abnormal index root causes are efficiently explained and traced.