251 results about "Preference data" patented technology

The invention discloses an advertisement putting strategy optimization method and system based on preference data collaboration, and relates to the technical field of advertisement putting, and the method comprises the steps: collecting the internal user data of a local advertisement putting platform, obtaining the external user data of a third-party platform through a safety data interface, carrying out the integration and preprocessing of the data, and obtaining an advertisement putting strategy; extracting explicit preference data and implicit preference data; a user-advertisement scoring matrix is constructed based on the preference data, and user preference vectors are generated after decomposition; acquiring copywriting text, image and category information of an advertisement to be put, constructing an advertisement feature expression model and generating an advertisement feature vector; calculating a matching degree score based on the user preference vector and the advertisement characteristic vector, and dynamically optimizing an advertisement putting strategy according to the score; according to the method, through comprehensive utilization of explicit and implicit preferences of the user, more accurate user interest modeling and advertisement matching are realized, and the advertisement putting effect and putting efficiency are effectively improved.

The invention relates to the technical field of video generation, in particular to an intelligent commodity video generation method based on multi-modal analysis and a dynamic narrative architecture, which comprises the following steps: inputting original commodity data and a user behavior log, and outputting a dynamic selling point weight vector through a selling point value evaluation engine; s2, inputting the dynamic selling point weight vector output in S1 into a narrative flow generator, and executing the following steps: intercepting first K core selling points according to the weight vector to form a narrative trunk chain; based on the historical preference data of the user, inserting an emotion enhancement node to generate a branch enhancement narrative flow; generating a multi-version narrative flow instruction set in combination with the real-time network bandwidth data; retrieving a preset video clip library according to the instruction set to generate an initial video sequence; detecting semantic fault regions between adjacent segments to generate compensation animation parameters; and injecting compensation animation parameters and outputting a continuous narrative video stream. According to the invention, the jamming feeling and the frame skipping phenomenon caused by the change of a narrative structure or different sources of video clips are reduced, and the overall watching experience of a user is improved.

The invention discloses an intelligent parking method and device based on multi-source data fusion and a medium, and relates to the technical field of big data technologies. The method comprises the following steps: collecting multi-source parking data, and analyzing the multi-source parking data to generate a parking lot label library; generating a user tag based on historical search records and residence time preferences of the user, and constructing a user information database; when it is detected that a user triggers a parking request, performing multi-dimensional matching calculation on a user tag and a parking lot tag library to obtain a recommended parking space sequence; and optimizing the recommended parking space sequence based on the real-time traffic state and the user credit score to generate a final navigation path. According to the method, a dynamic parking recommendation mechanism is constructed by collecting and analyzing the multi-source parking data and the user preference data.

The invention relates to a costume design system and method based on multi-modal AIGC and a storage medium thereof. Comprising a multi-source input processing module used for receiving and processing multiple modal inputs including costume design text description, a costume style drawing, a fabric sample image and user preference data; the clothing feature extraction module is used for extracting clothing structure features and visual style features from the multi-source input; the semantic understanding module is used for carrying out semantic understanding and style analysis on the clothing features; the multi-modal fusion module is used for mapping the clothing feature representations of different modals to a unified semantic space and generating a fusion feature vector; and the style migration module is used for realizing design generation of a specific style based on a small number of garment samples through a LoRA low-rank adaptation technology, and is used for generating a garment design drawing through a multi-stage conditional diffusion model based on the fusion feature vector.

The invention provides a reward model training method and device, a strategy model training method and device, electronic equipment and a storage medium, and belongs to the technical field of artificial intelligence, and the method comprises the steps: obtaining a preference data pair which comprises a preferred response and a non-preferred response generated for the same prompt word, and each of the preferred response and the non-preferred response is composed of a plurality of text unit sequences; inputting each text unit sequence into a to-be-trained reward model to obtain a predicted reward value; and calculating the total training loss according to the predicted reward value, and updating the model parameters of the to-be-trained reward model. According to the method, the response text is subjected to serialized splitting, and the preference data composed of the preferred response and the non-preferred response is introduced for comparative learning, so that the target of model training is no longer to evaluate the absolute quality of a single response, but to identify a key text unit which causes one response to be superior to the other response; the fine-grained evaluation of the response content is realized, and the evaluation accuracy of the reward model and the identification capability of complex user preferences are effectively improved.

The invention discloses an AI system for generating customized contents based on user preferences, which belongs to the technical field of artificial intelligence, and comprises a user preference modeling module for constructing and continuously updating a user feature vector library and realizing personalized demand modeling and classification; the interactive input and early warning module is used for receiving user input, implementing risk control and providing a visual interactive correction tool; an intelligent parameter conversion module; a video preprocessing and enhancing module; a cross-modal semantic generation module; the dynamic training and optimizing module is used for controlling a model training process and balancing repair quality and style migration; the multi-version generation and evaluation module is used for outputting a differentiated repair result and quantitatively evaluating the performance; the feedback learning and iteration module is used for collecting user preference data and driving the system to continuously optimize; and a distributed task scheduling module. According to the method, the real-time performance and accuracy of the user portrait are ensured through the layered modeling mode, accurate matching of personalized content generation is supported, and user experience and content recommendation are effectively improved.

The invention discloses a biochemical experiment automation script training generation method based on RAG and DPO. The method comprises the steps that biochemical experiment technical documents are collected to serve as an external knowledge base; the BM25 and the Fiss are fused to construct a hybrid retriever; generating experimental process description by using a large language model, determining equipment, protocols and materials, and retrieving related documents from a knowledge base on the basis of the experimental process description, the protocols and the materials; generating an experiment script based on a retrieval result, and performing platform simulation verification, and performing iterative optimization on the script which fails in verification; marking successful and failed scripts as preference data pairs, and constructing a training set; carrying out LoRA fine tuning on a local large language model by adopting direct preference optimization; and according to a target and appliance prompt input by a user, generating a verified experiment script by using the large language model after direct preference optimization training. According to the method, the script generation automation level and the passing rate are remarkably improved, the manual intervention cost is reduced, and an efficient solution is provided for automation of biochemical experiments.

The present disclosure provides a system for generating artist-governed merchandise derivative works, comprising at least one processor and memory storing instructions that, when executed, cause the system to receive a request associated with source content and a transformation theme for creating a merchandise derivative work, evaluate the request against pre-generation preference data associated with a content authority using a knowledge agent maintaining brand guidelines, generate the merchandise derivative work from the source content according to the transformation theme by applying a generative artificial-intelligence model through an action agent, perform post-generation validation against the brand guidelines, generate a digital identifier based on the merchandise derivative work, and enable distribution through a physical merchandise path for tangible products or a virtual merchandise path for digital assets.

The invention relates to the technical field of finance, and provides a time series data prediction method and system based on an auto-reflection mechanism large language model, and the method comprises the steps: firstly collecting structured transaction data and public unstructured text information in a preset historical time window, inputting a first large language model, and generating structured market background information in combination with a preset cue word; inputting the data and preset risk preference data into a second large language model to generate first causal reasoning training data; and inputting a third language model for reflection correction, checking whether the argument can be verified in the background information, and if so, reserving and strengthening to obtain second causal reasoning training data. And constructing a prediction large language model, optimizing parameters by means of a group relative strategy optimization algorithm, and then outputting the parameters. The system and the method depend on an auto-reflection mechanism, the model considers risk preferences in financial time sequence prediction to generate various risk strategies, manual annotation dependence is greatly reduced, the cost is reduced, and the risk preference module is independent and configurable and can be flexibly expanded.

The invention provides an optimization method, an interaction method and a system of a large language model. The optimization method comprises the steps that a preference data set is obtained, the preference data set comprises a plurality of preference pairs, one preference pair comprises question samples, preference answer samples and non-preference answer samples, the preference answer samples and the non-preference answer samples correspond to the question samples, and dynamic target reward difference values corresponding to all the preference pairs are determined, the dynamic target reward difference values corresponding to the at least two preference pairs are different, the large language model is optimized according to the dynamic target reward difference values, an optimized large language model is obtained, and the optimized large language model is used for determining a target answer meeting the preference of the target user according to the obtained target question of the target user. The optimized large language model not only can well distinguish preference answers from non-preference answers, but also can be finely adjusted according to different preference intensities, so that the personalized requirements of target users can be better met in practical application.

The invention discloses a direct preference optimization-based large model post-alignment training method and system. The method comprises the following steps of: (1) constructing a static reference model and a strategy model; (2) optimizing preference data by using an implicit reference model, and (3) adjusting a strategy model by optimizing an objective function so that the strategy model better fits the user preference. According to the method, the limitation of the existing DPO and SimPO methods is solved by introducing a new reference model expression, for example, the method depends on a suboptimal reference model or uses a fixed reward margin. By adjusting the balance between the strategy model and the reference model, personalized reference model setting for each pair of responses is realized, and meanwhile, a theoretical guarantee is provided and the KL divergence is effectively controlled. And the alignment performance and the winning rate of the model are remarkably improved, and the method is a robust LLM fine tuning method.

The invention relates to the technical field of Internet information processing, in particular to a sound effect recommendation method, system and device based on user preference and scene awareness, and the method comprises the steps: carrying out the deep analysis of collected user preference data, and constructing a user preference model; when an audio recommendation request of a user side is received, scene perception information matched with a current vehicle position is obtained, and a current scene is determined; matching the current scene with the user preference model to obtain a candidate sound effect set matched with the current scene and the user preference; checking each candidate sound effect data in the candidate sound effect set, screening the candidate sound effect data according to the sound effect data, and generating a recommendation result; and feeding back the recommended sound effect contained in the sound effect recommendation result. According to the scheme, the personalized requirements of the user can be fully considered, and the preferable sound effect recommendation information is provided for the user, so that the satisfaction degree of the user on sound effect recommendation is improved.

The invention relates to the technical field of internet advertisement marketing, in particular to an internet advertisement marketing method and system based on artificial intelligence, and the method comprises the steps: integrating the browsing history, demographic information and behavior preference data of a user, generating a dynamic user portrait, and predicting a user demand; generating personalized advertisement creative content based on the user portrait and the predicted user demand; performing multi-dimensional automatic violation review on the advertisement creative content, and performing originality verification on the advertisement creative content passing the violation review; based on the dynamic user portraits and the advertisement putting targets, advertisement creative contents passing violation review and originality verification are accurately pushed; according to the method, the advertisement can be automatically examined before the advertisement is put, so that the examination efficiency is improved, the compliance risk caused by omission due to manual detection and the brand reputation loss caused by the compliance risk are reduced, and the advertisement marketing effect can be improved.

A method for providing personalized non-profit venue visit recommendations to a visitor at a non-profit venue, comprising: providing a management interface for management of a set of multimedia assets; receiving a mapping of multimedia assets to a display in at least one site plan; receiving metadata for the set of multimedia assets; providing an interface for receiving data indicating personal interests of a visitor, including visitor-provided preference data and passively-collected visitor interaction data; receiving personal interest data for a plurality of non-profit venue visitors; receiving preference data relating to a prospective visitor; applying a machine learning system to analyze the metadata, the data indicating personal interests of the visitor, the personal interest data for the plurality of non-profit venue visitors, and the preference data relating to the prospective visitor; and generating a selection or sequence of non-profit venue location recommendations for the prospective visitor.

The invention discloses a cross-language migration-based low-resource programming language corpus enhancement method, which comprises the following steps of: receiving a source language code, and driving a large language model to generate an initial target language code through a two-way retrieval mechanism fusing example guidance and knowledge constraint; secondly, constructing an automatic iterative repair closed loop by utilizing the feedback of a compiler, and carrying out self-correction on codes which fail to be compiled; thirdly, high-quality codes are screened out through automatic quality gating and fed back to a corpus and a knowledge base, and self-enhancement circulation of data and knowledge is formed; finally, the method further comprises an offline model evolution step based on grammar and semantic alignment driven by a compiler, a big language model is trained by collecting preference data generated by an online process, and the ability of the big language model to understand a target language is improved fundamentally. According to the method, the core problems of low quality, poor efficiency and lack of self-evolution ability of a low-resource programming language in code migration are solved.

The invention relates to an intelligent decision-making method and system for server customization demand matching. The method comprises the steps of obtaining a server customization demand input by a user; constructing a multi-dimensional demand portrait based on the server customization demand, wherein the demand portrait comprises a quantitative index and a configuration constraint condition; according to the demand portrait, executing configuration compatibility check, and identifying whether a compatibility problem exists or not; when the compatibility check is passed, a candidate configuration scheme is generated by adopting a multi-target genetic algorithm based on the demand portrait passing the compatibility check; and sorting and screening the candidate configuration schemes based on user preference data in the demand portraits passing the compatibility check to obtain a recommended scheme list, and transmitting the recommended scheme list to a display interface for a user to select and confirm, the recommended scheme comprising a component list, a price, a performance index and a recommendation reason. The method has the effect of improving the accuracy of server customization recommendation.

The invention provides a promotion effect real-time analysis method based on multi-dimensional data, and the method comprises the steps: obtaining a shared content draft currently generated by a user through a camera application, recognizing the integrating degree of a core theme in the content and a platform content theme tendency based on a personalized text description template, generating a text description first draft conforming to a target platform style, and analyzing the promotion effect of the user according to the text description first draft. Outputting the optimized first version of the shared content; and if the user manually adjusts the first version of the shared content, recording the editing behavior details of the user, analyzing the relevance between the adjustment direction and the platform text format specification and the platform manifold use frequency, updating the platform tone style preference data in the user behavior feature library, and obtaining a user sharing habit data set.

The invention provides an electronic bidding document generation method and system based on artificial intelligence, and the method comprises the steps: firstly obtaining a bidding demand text issued by a bidding party, a historical bidding document material set accumulated by a bidding party, and bidding demand change records of similar projects in an industry; and constructing a bid inviting scene semantic map containing demand nodes, multi-modal material nodes, trend association nodes and inter-node dynamic semantic association strength information. Generating an electronic bidding document dynamic framework according to the bidding scene semantic map, and calling a cross-modal artificial intelligence content generation model to generate a multi-modal module content first draft of each function module unit; and then semantic consistency adjustment and connection optimization processing are carried out on the first draft of the multi-modal module content to obtain the integrated multi-modal electronic bidding document content. And finally, in combination with the presentation specification information and the industry bidding preference data, carrying out structural typesetting optimization and content adaptation adjustment processing on the integrated content to obtain a final multi-modal electronic bidding document, so that the quality and pertinence of the electronic bidding document are remarkably improved.

The invention provides a financial service recommendation method and device, equipment, a storage medium and a program product, and relates to the field of financial science and technology. The method comprises the steps of obtaining a user portrait label, behavior preference data of a user and a product portrait label; processing the behavior preference data of the user to obtain keyword label data representing the type characteristics of the financial service and / or the financial product, scoring the keyword label data, determining label integral data, and determining a grade score of the keyword label data based on the label integral data; and performing matching on the basis of the user portrait label and the product portrait label every preset duration, determining a target financial service and / or a target financial product, performing evaluation on the target financial service and / or the target financial product on the basis of the grade score, and determining a to-be-recommended financial service and / or a to-be-recommended financial product. According to the method, the recommendation accuracy and efficiency can be improved, and the customer satisfaction can be improved.

Systems, computer program products, and methods are described herein for dynamically configuring graphical user interface components based on interface interaction data. The present invention is configured to identify a user device associated with a user account; identify at least one user access to a platform from the user device; determine, by an emotional artificial intelligence (AI) engine, at least one user platform preference for the user account, wherein the emotional AI engine is pre-trained on historical user platform preference data for the user account; generate, by the emotional AI engine, a user platform interface component based on the at least one user platform preference; and transmit the user platform interface component to the user device, wherein the transmission of the user platform interface component triggers a configuration of the GUI of the user device.

The invention discloses a house rental information dynamic recommendation method in combination with real-time supply and demand data, and particularly relates to the technical field of information recommendation. The method comprises the following steps: acquiring original space boundary data to generate house rental boundary data; determining a supply and demand hot spot area by analyzing spatial distribution characteristics of supply-side house source data and demand-side user data; generating spatial overlapping relation data by analyzing the overlapping relation between different spatial scales; determining the influence range of region boundary overlapping by evaluating the crossing degree of supply and demand data; obtaining a corrected supply and demand hot spot area through spatial weight correction of the supply side house source data and the demand side user data; generating user recommendation preference data in combination with user historical recommendation feedback data; and recommending house renting information to the user according to the corrected supply and demand hot spot region and the user recommendation preference data. The data deviation caused by region boundary overlapping is effectively reduced, and the accuracy and individuation degree of house lease information recommendation are improved.

The invention provides a data generation model training method and device, electronic equipment and a storage medium, and belongs to the technical field of artificial intelligence, and the method comprises the steps: constructing a preference data pair which comprises an initial question and answer pair and an optimized question and answer pair; inputting the preference data pair into a to-be-trained large language model, and obtaining an enhanced question and answer pair output by the to-be-trained large language model; updating the preference data pair to obtain an updated preference data pair, including: setting the enhanced question and answer pair as an optimized question and answer pair in the updated preference data pair; and iteratively executing the steps of constructing the preference data pair and updating the preference data pair until a preset termination condition is met, and obtaining a trained large language model. According to the method, the question and answer pair data is optimized by utilizing the ability of the large language model, and the large language model is trained by using the preference data pairs subjected to iterative updating, so that high-quality question and answer pair data can be generated efficiently at low cost.

The invention discloses a human preference alignment method based on a self-improvement large visual language model. The method comprises the following steps: automatically constructing a preference data set; acquiring an image-question pair from the visual question and answer data set; applying various visual enhancements to the image, and driving a reference model to generate a group of candidate answers in combination with a question; the group of candidate answers are summarized and extracted into a more comprehensive'win 'text, and meanwhile, the text answers of the reference model to the original picture and the question are taken as'fall-fail' texts, so that preference data in an'image-inquiry-'win 'text-'fall-fail' text 'format are formed; secondly, based on the preference data set constructed in the first step, a direct preference optimization algorithm is applied to conduct alignment fine adjustment on a target visual language model, in the fine adjustment process, parameters of a visual encoder are kept frozen, and a low-rank adaptation layer (LoRA) is only introduced into an extended mode alignment module and a language decoder for training; carrying out iterative self-improvement on the model; after fine tuning is completed, taking the optimized model as a new reference model, and repeating the data construction process in the step 1 to generate preference data with higher quality for the next round of optimization; and the circulation is repeated, so that the continuous self-improvement of the model alignment capability is realized. According to the method, self-supervised preference alignment without manual annotation is realized, and a more comprehensive and high-quality'win 'text is generated.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for more efficiently configuring a policy model to generate candidate messages. One of the methods includes prompting a policy model to generate candidate messages for new content being introduced to the system in reference to a control message associated with the new content. A reward model predict a performance of at least one of the candidate messages for the new content against the control message associated with the new content. The candidate messages are tested to obtain actual relative preference data obtained from engagements with the candidates messages being tested. The phantom relative preference data are supplemented with real relative preference data. Candidate messages are selected to send as continuing text.

A data processing system implements a framework that utilizes in-situ user interactions as a source of feedback for improving the training of LLMs to generate outputs that align with user preferences. The framework includes a user preference evaluation pipeline analyzes in-situ user interactions with the LLMs and generates preference information that can be used to improve the training of the LLM to improve the alignment of the LLM with user preferences. The user preference evaluation pipeline includes a feedback signal identification unit that identifies explicit and / or implicit feedback provided by the users in response to content output by the LLM in response to a user prompt. The feedback signal identification unit estimates user satisfaction with a set of satisfaction rubrics and user dissatisfaction with a set of user dissatisfaction rubrics to generate user preference data that can be used to align an LLM with these user preferences.

The invention provides a hotel environment adaptive regulation and control system based on the Internet of Things, and belongs to the field of hotel regulation and control management. Comprising a privacy enhanced environment perception and preference acquisition module, a hotel operation data security fusion engine, a dynamic room state and authority regulation and control decision module and a security execution and auditing module which perform data interaction to enhance user data confidentiality and deeply align with hotel operation information; the privacy enhanced environment perception and preference acquisition module is deployed in a guest room and a public area and comprises an environment sensor and a user interaction terminal; a preference acquisition mechanism based on a temporary anonymous identifier is adopted, a user activates an interactive terminal through a check-in voucher and establishes an encrypted anonymous identifier uniquely corresponding to a physical room number, and acquired environment setting preference data is only bound with the anonymous identifier and is stored in an encrypted storage area of a local edge computing node of a guest room.

The invention discloses an interactive advertisement engine implementation method and device, a server and a storage medium, and belongs to the technical field of advertisement technology and intelligent interaction, and the method comprises the steps: generating advertisement content containing interaction elements based on advertisement materials and pre-collected user preference data; in the playing process of the advertisement content containing the interaction elements, a voice instruction of a user is received and analyzed, and corresponding voice interaction response is controlled to be carried out on the interaction elements in the advertisement content according to a voice instruction analysis result; in the playing process of the advertisement content containing the interaction elements, analyzing a gesture interaction instruction of a user when the gesture interaction instruction is received, and controlling the interaction elements in the advertisement content to perform corresponding gesture interaction response according to the analysis result of the gesture interaction instruction; and optimizing an advertisement generation strategy based on the analyzed interaction data of the user. According to the invention, interaction between the user and the advertisement content through voice or gestures is supported, so that the attraction and effect of the advertisement are improved.

The invention discloses a multi-agent collaboration system for constructing a data set for vertical field large language model training. The system integrates four levels of data acquisition, data processing, data generation and data optimization, and can automatically complete full-process processing from multi-source corpus acquisition, structured analysis to instruction and preference data construction after receiving a field demand and a retrieval intention proposed by a user. Through the cooperative effect of picking, structured reconstruction, co-citation analysis, expert generation and screening optimization intelligent agents, automatic construction of a pre-training data set, an instruction fine tuning data set and human feedback reinforcement learning RLHF data is realized, and the consistency, traceability and high quality of the data are ensured; therefore, the training effect and the application performance of the vertical field large language model are effectively improved.

The application discloses a KTV system rear surround setting method and device, a terminal and a medium. The method comprises the following steps: acquiring rear surround device static parameters, box environment data, box acoustic data and customer preference data; determining a target position of the rear surround device in the box based on the rear surround device static parameters, the box environment data and the box acoustic data; determining a target adjustment parameter combination corresponding to the rear surround device based on the rear surround device static parameters, the box acoustic data and the customer preference data; processing the rear surround device based on the target adjustment parameter combination, and placing the rear surround device at the target position. The application aims to achieve the optimal acoustic effect of the rear surround device of the KTV system in the box, and ensure that the acoustic effect of the rear surround device meets the preferences of customers in the box.

The invention discloses a large language model alignment method based on multi-user preference dynamic balance, and the method comprises the steps: collecting user interaction data, preference comparison data and public policy question and answer data in a livelihood consultation scene, and dividing the data into structured preference data and unstructured data; preprocessing the divided data to generate a user feature vector and a response feature vector; constructing a user-response bipartite graph, performing multi-hop preference propagation through a graph neural network, capturing potential association among users, and outputting a user preference embedding vector; embedding a vector selection expert path according to user preference through a dynamic gating function, and generating a personalized response; and through an optimization-free embedding aggregation strategy, similar users are retrieved based on a graph structure, and the embedding of the similar users is weighted and aggregated, so that rapid cold start of new users is realized. The precision and efficiency of livelihood consultation can be remarkably improved, and the method is particularly suitable for smart city scenes with high-frequency policy updating and frequent user flow.