13results about How to "Reproducible" patented technology

The invention provides a large language model text traceability method based on virtual cue word embedding, and belongs to the technical field of artificial intelligence natural language processing and the technical field of digital content security, the method comprises the following steps: (1) constructing a chat input instruction text with a role format; (2) obtaining embedded vector input; (3) obtaining system cue word embedding representation; (4) obtaining a generated text with a user identity information watermark; and (5) recovering user identity information to realize traceability of the generated text. According to the method, on the premise that main body parameters of the large language model are not modified and model text generation quality and response efficiency are not affected, instant embedding of traceable identity information in the large language model generation process is achieved, meanwhile, robustness of embedded watermark information in a text transformation scene is guaranteed, and the method is suitable for large language model generation. Verifiable and reproducible identity identification information can be stably extracted from the generated text, and reliable technical support is provided for source tracing of the generated text.

This invention provides a method for tracing the source of large language model text based on virtual prompt word embedding, belonging to the fields of artificial intelligence natural language processing technology and digital content security technology. The method includes: (1) constructing chat input command text with role format; (2) obtaining embedded vector input; (3) obtaining system prompt word embedding representation; (4) obtaining generated text with user identity information watermark; (5) restoring user identity information to achieve source tracing of generated text. This invention achieves real-time embedding of traceable identity information in the large language model generation process without modifying the main parameters of the large language model or affecting the model text generation quality and response efficiency. At the same time, it ensures the robustness of the embedded watermark information in text transformation scenarios, and can stably extract verifiable and reproducible identity information from the generated text, providing reliable technical support for tracing the source of generated text.

The invention provides a method, a device, equipment and a medium for implementing financial domain data governance and data-in-platform cooperation, and the method comprises the steps: constructing an enterprise-level data governance system, and determining a data management framework and a standard specification; based on a full-stack information creation technology stack, deploying a data middle platform supporting a lake-warehouse integrated framework to form a unified data storage, calculation and governance capability core; injecting an output rule of the data governance system into a data platform by taking the financial business domain as a governance template, governing the financial domain data, and generating financial data assets and services; the managed financial data assets are packaged into data services through the data middle platform, and operation of a financial sharing service system and financial data analysis application is supported; in the application process of financial data assets, feedback information is collected, rules of a data governance system and a governance flow of a data-in-platform are dynamically optimized, and closed-loop control is formed, so that the problems that the data governance system and the data-in-platform are disjointed, and a data-in-platform solution always faces challenges are solved.

The embodiment of the invention provides a causal path completion method and device, equipment and a storage medium, and relates to the technical field of network security. In response to the path completion instruction, acquiring a path breaking point from the causal path, and determining a context data packet according to the path breaking point; performing large language model reasoning on the context data packet based on a vector database to obtain verification plan data of the candidate path, and arranging the verification plan data according to the verification platform information to obtain an instruction sequence containing at least one atomic action instruction; performing verification on the instruction sequence to obtain a verification result corresponding to each atomic action instruction; and performing decision judgment on the candidate path according to the verification plan data and the verification result to obtain a judgment result, and complementing the causal path based on the judgment result. The whole process from fracture point positioning, candidate path reasoning to verification completion is automatic, the processing efficiency is effectively improved, and the accuracy and credibility of a completion result are guaranteed.

The application discloses a kind of intelligent detection building health and risk prediction teaching system, belong to intelligent construction teaching field, including: through intelligent car and the high-definition camera matrix of unmanned aerial vehicle terminal carrying is collected building data and is transmitted to operation super brain;Through the AI recognition big model and structure safety, natural disaster, personnel safety three risk assessment sub-models built-in in operation super brain, the multi-level comprehensive risk assessment of building state is carried out;Through the hazard identified in building digital twin 3D model corresponding position visualization presentation and generating graded warning signal by wisdom management platform, platform supports student login operation to carry out picture digitization, autonomous operation practice and complete experiment report;After completing on-campus laboratory simulation and course practice, form course outline and experiment report and popularize to other colleges and universities.The application enables students to visually observe building hazards, complete risk prediction operations, and promote cross-campus sharing of teaching resources.

The application relates to a single-molecule transistor structure coupled with a THz wave band and a preparation method thereof. The single-molecule transistor structure comprises a silicon substrate, a gate, a gate dielectric layer, a source and a drain matched with the gate control, a split ring resonator structure, a peripheral electrode and a single-molecule active region located at a nanometer / sub-nanometer gap of the source and the drain. Through the structure and the method, a significant terahertz local field enhancement can be obtained in the gap region where the molecules are located while the transistor characteristics and the device stability are ensured, and an implementable process path is provided for the construction and application research of the single-molecule coupling device in the terahertz wave band.

The invention relates to a photovoltaic unmanned aerial vehicle cleaning path candidate generation method based on representative sampling, and the method comprises the steps: designing a small number of balanced and low-correlation factor combinations around key factors such as starting point selection, an intra-cluster construction strategy, a weight-distance tradeoff coefficient and a cross-grade connection rule through a small-sample representative combination sampling method, and carrying out the selection of the cleaning path candidate of a photovoltaic unmanned aerial vehicle. Quickly generating a small batch of array cluster access sequence candidates with strong representativeness; under each cluster sequence, a component-level path is generated according to a set intra-cluster component sequence and direction optimization strategy; and finally, combining the clusters with the paths in the clusters to generate all candidate paths. According to the method, dirt weight rapid coverage and flight path / time cost control are taken into consideration, and high-quality candidate paths are provided for subsequent global path scoring and screening.

The invention relates to the technical field of biological information, in particular to a generic cancer brain metastasis molecular typing system and method and a medium. According to the generic cancer brain metastasis molecular typing system, brain metastasis molecular typing obtained through a data processing and consensus clustering method is kept stable in multiple queues and multiple platforms, and reproducibility and robustness are achieved. The typing system disclosed by the invention can identify four subtypes with consistent biological connotation in general cancer brain metastasis, can realize unified brain metastasis typing across primary parts, and avoids the limitation of typing only depending on primary lesion. The four subtypes correspond to different states such as neuron-like, immune infiltration, metabolic reprogramming and proliferation procedures and can be consistently verified on the proteome / pathway level, and biological interpretation is high. The four types can correspond to prognosis differences, immune / metabolic landscapes and other characteristics, and a basis is provided for risk stratification and potential treatment strategy selection.

This invention provides a method, apparatus, and electronic device for improving the fluidity of porous media using pulsating flow. The method includes: determining a critical threshold and a subcritical baseline; applying a steady-state displacement with a displacement intensity equal to the subcritical baseline, and superimposing a pulsating drive; using a solution framework to advance the interfacial motion of droplets under pulsating drive over time, adaptively controlling the time step during the advancement process, the time step being jointly limited by multiple physical constraints; determining the truncated axial net displacement of the droplet within the pore throat contraction structure within a fixed observation time window; obtaining the fluidity index corresponding to the pulsating drive based on the truncated axial net displacement; keeping the amplitude of the pulsating drive constant, changing the frequency of the pulsating drive to obtain a first curve of the fluidity index changing with frequency, and determining a cutoff frequency based on the first curve; changing the amplitude of the pulsating drive to obtain a second curve of the cutoff frequency changing with amplitude. This method is suitable for engineering parameter design and selection and is reproducible.

The application discloses a BIM-based digital twin model creation method, comprising: S1 semantic fingerprint evaluation and completion, calculating a semantic fingerprint vector from three dimensions of geometric integrity, attribute integrity and relationship integrity to identify and complete an unready component; S2 intention graph clipping and enhancement, querying an intention graph according to a target application scenario to clip a minimized submodel and performing directional semantic enhancement; S3 semantic geometry decoupling construction, extracting a building topology skeleton to construct a semantic topology layer and a geometric performance layer respectively, and outputting a decoupled multi-resolution digital twin model; and S4 function semantic inference injection, inferring implicit function information based on a function space semantic graph and a building function knowledge base and injecting the semantic topology layer. The application improves the data quality, scene adaptability and semantic depth of the digital twin model.

This application discloses an optimized deployment method and system for traveling wave location devices in complex multi-branch distribution networks. It transforms the topology of a radial distribution network into a tree-like hierarchical model, establishes a linear constraint system for observable faults across the entire network based on the adjacency relationship between parent and child nodes, and then constructs a zero-one-integer programming model with the objective of minimizing the total number of devices installed. A weighted adjustment of the objective function is made by introducing a combined weight of line length and historical fault frequency. Finally, an accurate solution is achieved through a mixed-integer linear programming solver. This approach significantly reduces the number of traveling wave location devices required, substantially lowers equipment investment costs, and ensures that the optimization results are globally optimal and reproducible. Furthermore, the weighted mechanism ensures that the deployment scheme is highly adaptable to the actual operating conditions of the distribution network, effectively enhancing the engineering practical value of fault location.

This invention relates to a ternary heterojunction catalyst electrode, its preparation method, and its application in the electrocatalytic degradation of dyes. The preparation method of the ternary heterojunction catalyst electrode of this invention involves: dissolving a sulfur source in an MXene dispersion, heating and stirring to obtain a first mixture; dissolving a molybdenum source in the first mixture, heating and stirring to obtain a second mixture; immersing nickel foam in the second mixture and placing it in a reactor for heating and reaction to obtain the ternary heterojunction catalyst electrode. The ternary heterojunction catalyst electrode uses nickel foam as a substrate, the surface of which is coated with MXene, and the surface of the MXene is supported with intercalated molybdenum disulfide; wherein the MXene and molybdenum disulfide are bonded through Ti-O-Mo bonds. This invention, by constructing a heterojunction and utilizing interface engineering to synergistically optimize the electronic structure, can effectively enhance the electron transfer rate, achieving a highly efficient electrocatalytic activity of "1+1>2".

The application provides a kind of power grid environmental protection monitoring ability evaluation method and device, belongs to environmental protection monitoring technical field.The method includes: obtaining the score data that multiple experts score the importance degree and overall importance degree of each environmental protection monitoring ability evaluation index, according to the score data, establish initial matrix;The covariance matrix of initial matrix is calculated, and according to each eigenvalue of covariance matrix, the corresponding eigenvector is selected to establish feature matrix;According to the feature matrix, the dimensionality of initial matrix is reduced to obtain the final matrix;According to the final matrix, the weight of each environmental protection monitoring ability evaluation index is calculated using the preset linear regression algorithm;Based on the weight and the historical data corresponding to each environmental protection monitoring ability evaluation index in the power grid project to be evaluated, the environmental protection monitoring ability evaluation result of the power grid project to be evaluated is determined.The environmental protection monitoring ability of power grid enterprise or project can be objectively and accurately evaluated.