An advertisement material searching method and system based on quantum heuristic algorithm

Abstract

This invention discloses a method and system for searching advertising creatives based on quantum heuristic algorithms. The method specifically includes: encoding visual content features into a principal quantum state based on an advertising creative feature set, and encoding audio features, text semantic features, and user preference features into auxiliary quantum states coupled to the principal quantum state, forming a cross-modal quantum encoding result; dynamically updating the encoding primitives, coupling weights, and constraint labels in the cross-modal quantum encoding result to generate a dynamic quantum encoding library; constructing a QUBO model based on the dynamic quantum encoding library; solving the QUBO model using quantum annealing or a quantum heuristic algorithm to obtain a target advertising creative combination; decoding and matching the target advertising creative combination, and generating corresponding advertising delivery plans according to differences in terminal environments. This invention improves advertising creative search efficiency, combination optimization capabilities, real-time operational adaptability, and cross-platform advertising effectiveness.

Description

Technical Field

[0001] This invention relates to the field of intelligent search and delivery optimization technology for advertising creatives, and in particular to an advertising creative search method and system based on quantum heuristic algorithms. Background Technology

[0002] MMO game advertising creatives typically include various modal elements such as character images, class settings, skill effects, scene atmosphere, event themes, background music, and copywriting selling points. With the ever-accelerating pace of game updates, holiday events, cross-server battles, new dungeons, and new class releases, advertising creatives need to be quickly adjusted and precisely targeted based on real-time events and target users.

[0003] Currently, common methods for searching advertising creatives mainly rely on keyword retrieval, manual tag matching, or recommendation models based on historical click-through rate (CTR) and conversion rate (CVR). However, these existing methods have the following shortcomings in practical applications: (1) Difficulty in integrating multimodal material features. Visual materials, audio materials, text semantics and in-game behavior data are usually processed separately and lack a unified feature expression method. As a result, material search cannot simultaneously take into account multi-dimensional information such as visual impact, auditory emotion, text selling points and user preferences, which affects the accuracy and comprehensiveness of material matching.

[0004] (2) Difficulty in responding to real-time operational changes in MMO games. Operational statuses such as version updates, holiday events, cross-server battles, and changes in server load directly affect advertising strategies and material performance. However, traditional methods cannot promptly feed this dynamic operational information back into the material search model, resulting in insufficient timeliness of search results and a disconnect from current operational activities.

[0005] (3) Difficulty in handling complex material combination search space. MMO advertising material combinations are extremely high in dimensionality. There are a huge number of possible combinations among roles, professions, skills, scenes, music, copywriting and advertising platforms. Traditional exhaustive search or rule-based matching methods are inefficient in dealing with such high-dimensional combination optimization problems and it is difficult to find the optimal or near-optimal material combination within an acceptable time.

[0006] (4) Lack of joint evaluation of user emotions and in-game behavior. Traditional metrics such as CTR and CVR cannot fully reflect users’ true emotions and interests towards characters, world view, event benefits and social gameplay. At the same time, they ignore the impact of material fatigue, negative user feedback and cross-modal content consistency on user experience, resulting in material selection that cannot accurately match players’ deep preferences.

[0007] (5) Inconsistent advertising experience across platforms. There are significant differences in user preferences, screen ratios, interaction methods and platform material specifications among PC, mobile and console terminals. Existing methods are difficult to balance maintaining brand experience consistency and achieving platform differentiation adaptation, which can easily lead to distorted display of advertising materials, slow loading or fragmented experience, and reduce advertising effectiveness. Summary of the Invention

[0008] The purpose of this invention is to provide an advertising creative search method and system based on quantum heuristic algorithms. By uniformly encoding visual, audio, text, event, operational, and user behavior data into a quantum heuristic optimizable representation and constructing a QUBO model with dynamic constraints, the invention improves the efficiency of advertising creative search, combinatorial optimization capabilities, real-time operational adaptability, and cross-platform advertising effectiveness, thereby solving at least one of the aforementioned problems in the prior art.

[0009] In a first aspect, the present invention provides an advertising creative search method based on a quantum heuristic algorithm, the method specifically comprising: Acquire multimodal data of game content, real-time operational data, user feedback data, and terminal environment data of MMO games, and generate advertising material feature sets based on unified feature specifications; Based on the feature set of advertising materials, visual content features are encoded into the main quantum state, and audio features, text semantic features and user preference features are encoded into auxiliary quantum states coupled with the main quantum state, forming a cross-modal quantum coding result; Based on changes in game version, event theme, and operational status, the encoding primitives, coupling weights, and constraint labels in the cross-modal quantum coding results are dynamically updated to generate a dynamic quantum coding library. Based on a dynamic quantum coding library, a QUBO model is constructed with advertising effectiveness and experience consistency as optimization objectives and operational capacity and platform adaptation as dynamic constraints. The QUBO model is solved using quantum annealing or quantum heuristic algorithms to obtain the target combination of advertising creatives; The system decodes and matches the target ad creative mix, and generates corresponding ad delivery plans based on differences in terminal environments.

[0010] Secondly, the present invention provides an advertising material search system based on a quantum heuristic algorithm, the system specifically comprising: The data acquisition module is used to acquire multimodal data of game content, real-time operation data, user feedback data and terminal environment data of MMO games, and generate advertising material feature sets based on unified feature specifications; The quantum coding module is used to encode visual content features into the main quantum state based on the feature set of advertising materials, and encode audio features, text semantic features and user preference features into auxiliary quantum states coupled with the main quantum state, forming a cross-modal quantum coding result; The dynamic update module is used to dynamically update the encoding primitives, coupling weights, and constraint labels in the cross-modal quantum coding results based on changes in game version, event theme, and operational status, thereby generating a dynamic quantum coding library. The model building module is used to build a QUBO model based on a dynamic quantum coding library, with advertising effectiveness and experience consistency as optimization goals and operational capacity and platform adaptation as dynamic constraints. The model solving module is used to solve the QUBO model using quantum annealing or quantum heuristic algorithms to obtain the target ad creative combination; The ad delivery module is used to decode and match target ad creative combinations, and generate corresponding ad delivery plans based on differences in terminal environments.

[0011] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, and a computer program stored in the memory, wherein when the computer program is executed on the processor, it implements the advertising material search method based on a quantum heuristic algorithm as described in any of the above methods.

[0012] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the advertising material search method based on a quantum heuristic algorithm as described in any of the above methods.

[0013] Compared with the prior art, the present invention has at least one of the following technical effects: 1. By uniformly encoding visual, audio, text, event, operational, and user behavior data into a quantum-inspired optimizable representation and constructing a QUBO model with dynamic constraints, we can improve the efficiency of ad creative search, combination optimization capabilities, real-time operational adaptation capabilities, and cross-platform ad performance.

[0014] 2. Achieve unified expression and fusion search of multimodal material features, enabling material search to simultaneously integrate visual, auditory, semantic, and user preference information, overcoming the shortcomings of multimodal information separation in traditional methods, and significantly improving the comprehensiveness and accuracy of material matching.

[0015] 3. Enables material search to respond in real time to changes in MMO game operations, thereby quickly feeding back real-time operational information such as server capacity, server status, and event duration to the material search model, ensuring that the searched materials are highly relevant to the current operational strategy, and significantly improving the timeliness of materials and the speed of operational response.

[0016] 4. By constructing a QUBO model with advertising effectiveness and experience consistency as optimization goals and operational capacity and platform adaptation as dynamic constraints, candidate variables of creative materials, effect weights, experience weights, constraint penalties, and the synergistic and conflict relationships between creative materials are uniformly incorporated into a quadratic unconstrained binary optimization structure. Quantum annealing or quantum heuristic algorithms are used for solving the problem, which can quickly approach the global optimal solution in a large-scale creative combination space, significantly improving search efficiency and solution quality.

[0017] 5. When generating performance evaluation weights and experience consistency weights, we not only utilize historical performance data such as exposure, clicks, conversions, retention, and monetization, but also integrate negative feedback, creative fatigue, activity matching data, and consistency evaluations between visual, audio, text, and preferences. This forms a comprehensive portrayal of users' deep interests and emotions, making the search results more aligned with players' real interests in characters, worldviews, event benefits, and social gameplay, thereby enhancing ad appeal and user satisfaction.

[0018] 6. When generating advertising campaign plans, based on differences in terminal operating systems, screen ratios, resolutions, network conditions, device performance, and platform material specifications, targeted adaptations are made to the display ratio, material format, video bitrate, copy language, and landing page path of the target advertising materials. This achieves precise platform-differentiated delivery while maintaining brand visual and experiential consistency, thereby improving the consistency of cross-platform advertising experience and increasing advertising effectiveness and resource utilization. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating an advertising material search method based on a quantum heuristic algorithm provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an advertising material search system based on a quantum heuristic algorithm provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0021] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0022] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0023] In this application embodiment, the entity executing the process includes a terminal device. This terminal device includes, but is not limited to, devices capable of executing the methods disclosed in this application, such as servers, computers, smartphones, and tablets. Figure 1 This diagram illustrates a flowchart of an advertising creative search method based on a quantum heuristic algorithm, as disclosed in an embodiment of the present invention. The advertising creatives for the MMO game include character image materials, class display materials, skill effect videos, game scene screenshots, holiday or version event theme materials, background music, advertising copy, landing page materials, and display components required for different platforms. By uniformly encoding, dynamically updating, and optimizing the combination of the above multimodal materials, an advertising delivery scheme adapted to different terminal environments is generated, as detailed below: S101 acquires multimodal data of game content, real-time operation data, user feedback data, and terminal environment data of MMO games, and generates a feature set of advertising materials based on unified feature specifications.

[0024] In this embodiment, the system obtains basic data from the game material library, advertising material library, operation activity system, user behavior system, and advertising platform. The game content multimodal data includes character illustrations, class settings, skill demonstration videos, dungeon scenes, battle scenes, background music, sound effects clips, ad titles, ad text, and activity descriptions; real-time operation data includes the current game version number, holiday event status, new class release status, cross-server battle status, new dungeon opening status, server load status, and remaining event time; user feedback data includes user clicks, viewing time, conversions, payments, comment sentiment, bounce rates, blocking, complaints, and material fatigue data; terminal environment data includes terminal type, operating system, screen ratio, resolution, network status, device performance level, ad placement size, and platform material specifications.

[0025] After cleaning, deduplicating, and standardizing the format of the aforementioned data, the system extracts advertising material features according to a unified feature specification. For example, character appearance, profession style, skill colors, and scene atmosphere are converted into visual features; background music rhythm, emotional intensity, and timbre type are converted into audio features; welfare points, gameplay points, social points, and combat power growth points in the copy are converted into textual semantic features; and user preferences for professions, factions, combat gameplay, social gameplay, and welfare activities are converted into user preference features. This forms an advertising material feature set that can be used for subsequent quantum encoding.

[0026] S102, based on the feature set of advertising materials, encodes the visual content features into the main quantum state, and encodes the audio features, text semantic features and user preference features into auxiliary quantum states coupled with the main quantum state, forming a cross-modal quantum coding result.

[0027] In this embodiment, the system uses the visual content of the advertising material as the primary expression object, encoding visual content features such as character design, professional identity, skill effects, scene color scheme, and visual impact into a primary quantum state. This primary quantum state characterizes the core appeal of the advertising material in the user's first visual perception. Subsequently, the system encodes audio features, textual semantic features, and user preference features into auxiliary quantum states, establishing coupling relationships with the primary quantum state. For example, when a character's skill video is accompanied by high-energy background music, emphasizes the tagline "cross-server battle," and the target user has historically shown a high interest in competitive gameplay, the system increases the coupling strength between the visual material and the corresponding audio, tagline, and user preferences. Conversely, when the visual atmosphere is inconsistent with the music's mood, or when the tagline's selling points differ significantly from user preferences, the system decreases the corresponding coupling strength. Through this method, the system forms a cross-modal quantum encoding result that can simultaneously characterize visual, auditory, semantic, and user interests.

[0028] S103 dynamically updates the encoding primitives, coupling weights, and constraint labels in the cross-modal quantum coding results based on changes in game version, event theme, and operational status, generating a dynamic quantum coding library.

[0029] In this embodiment, when the game undergoes a version update, a new class is launched, a holiday event begins, a cross-server battle is initiated, or a new dungeon is opened, the system receives real-time status information sent by the operations system and adjusts the encoded content of the advertising materials according to this real-time status information. For example, during the launch of a new class, the system increases the weight of character images, skill effects, class descriptions, and trial landing pages related to the new class; during holiday events, the system increases the relevance of holiday scenes, limited-time benefits, festive music, and event descriptions; when the server load of a certain region is high, the system reduces the priority of high-traffic advertising materials targeting that region, or adds traffic diversion constraint tags.

[0030] Meanwhile, the system updates the constraint tags in the encoding results based on the advertising platform's review rules, creative validity period, budget consumption status, changes in user negative feedback, and changes in creative fatigue. For example, restriction tags are added to creatives that have failed review, expired campaigns, are inaccessible, or have excessively high user negative feedback; priority tags are added to creatives that perform stably and are highly relevant to the current campaign. The updated encoding results are written into a dynamic quantum encoding library to ensure that creative searches can be adjusted in real time according to changes in the game's operational status.

[0031] S104, based on a dynamic quantum coding library, constructs a QUBO model with advertising effectiveness and experience consistency as optimization objectives and operational capacity and platform adaptation as dynamic constraints.

[0032] In this embodiment, the system establishes candidate variables for each candidate advertising creative based on the creative records in the dynamic quantum coding library, and maps these candidate variables to creative identifiers, creative types, campaign tags, delivery platforms, ad placement specifications, and resource addresses. The system optimizes the consistency between advertising effectiveness and user experience based on expected click-through rates, conversion rates, user preference matching, campaign theme matching, and brand performance consistency. Simultaneously, it considers creative fatigue, user negative feedback risk, content conflict risk, and cross-modal inconsistency risk as factors that reduce selection priority.

[0033] Furthermore, the system incorporates operational capacity and platform compatibility requirements as dynamic constraints. For example, the system limits the number of creatives, budget range, exposure frequency, server capacity, campaign validity period, ad placement size, creative format, video duration, bitrate range, and platform review status within the same advertising campaign. For creatives with complementary relationships between roles, professions, skills, scenes, music, and copy, the system establishes collaborative relationships to encourage their joint selection; for creatives with duplicate copy, conflicting visual styles, inconsistent campaign versions, excessive overlap in target audiences, or conflicting platform rules, the system establishes conflicting relationships to inhibit their simultaneous selection. Thus, the system combines creative candidate variables, optimization objectives, dynamic constraints, and collaborative or conflicting relationships between creatives into a QUBO (Quadratic Unconstrained Binary Optimization) model.

[0034] S105 uses quantum annealing or quantum heuristic algorithms to solve the QUBO model and obtain the target advertising material combination.

[0035] In this embodiment, the system selects the solution method based on current computing resources. When a quantum annealing device or quantum annealing cloud service is available, the system converts the QUBO model into an input structure recognizable by the quantum annealing device and sets the number of samplings, annealing time, and solution rounds to obtain multiple candidate material combinations. When quantum annealing resources are unavailable or a fast local response is required, the system uses quantum heuristic algorithms for solving the problem, such as simulated annealing, parallel local search, or other heuristic optimization methods, to search for an approximate optimal solution in the high-dimensional material combination space.

[0036] The system performs constraint verification and sorting on the multiple candidate results obtained from the solution. Specifically, the system prioritizes retaining candidate combinations with higher predicted advertising performance, better cross-modal experience consistency, higher degree of operational constraint satisfaction, and lower platform adaptation risk. For candidate combinations that violate strong constraints, such as those containing expired campaign materials, materials that failed review, materials that exceed ad slot specifications, or materials that pose a risk to server capacity, the system will remove or replace them. Finally, the system determines one or more target ad creative combinations.

[0037] S106 decodes and matches the target ad creative combination, and generates corresponding ad delivery plans according to the differences in terminal environment.

[0038] In this embodiment, the system decodes the variable results in the target advertising material combination into specific advertising material identifiers, material versions, material types, resource addresses, delivery channels, and applicable ad placements based on the mapping relationships in the dynamic quantum coding library. Subsequently, the system re-matches and verifies the decoded materials according to the current delivery task's activity theme, target user group, delivery time window, and platform specifications, eliminating materials with inaccessible resources, inconsistent versions, or that do not meet review requirements, and replacing them with alternative materials when necessary.

[0039] Furthermore, the system reads terminal environment data and generates advertising delivery plans based on terminal differences. For example, for high-performance mobile terminals, the system generates a delivery plan that includes high-definition vertical videos, dynamic skill effects, and immersive landing pages; for weak network environments, the system generates low-bitrate short videos or static image plans; for PC widescreen ad slots, the system generates horizontal scene displays and multi-role combination copywriting plans; for console or large-screen environments, the system generates high-definition scene atmosphere display plans. The system also adjusts the display ratio, video duration, copywriting length, button style, landing page path, exposure frequency, and budget allocation according to the material specifications of different platforms.

[0040] Through the above implementation methods, this application can uniformly express the visual, audio, textual semantics, and user preferences in MMO game advertising materials, and can dynamically adjust the material encoding and search conditions in conjunction with version updates, activity changes, and operational status. Simultaneously, by using the QUBO model and quantum annealing or quantum heuristic algorithms to handle the high-dimensional material combination optimization problem, a target advertising material combination that meets the requirements of advertising effectiveness, experience consistency, operational capacity, and platform adaptation can be obtained in a relatively short time. Therefore, this application can improve the accuracy, real-time performance, and cross-platform delivery consistency of MMO game advertising material searches.

[0041] In some embodiments, step S101 above, which involves acquiring multimodal data of MMO game content, real-time operational data, user feedback data, and terminal environment data, and generating an advertising material feature set based on a unified feature specification, specifically includes: Establish unified feature specifications in advance, and set material identifiers, activity identifiers, region / server identifiers, user group identifiers, and terminal platform identifiers through unified feature specifications; Acquire multimodal data of game content, real-time operational data, user feedback data, and terminal environment data, and associate and bind them based on material identifiers, event identifiers, server identifiers, user group identifiers, and terminal platform identifiers to form material data records; Data preprocessing and feature normalization are performed on the material data records to obtain standardized material features; In accordance with unified feature specifications, standardized material features are organized into advertising material feature sets.

[0042] In this embodiment, the system pre-establishes a unified feature specification. This unified feature specification enables data from the game asset library, operation system, user behavior system, and advertising platform to be organized and accessed according to the same data caliber. Specifically, the unified feature specification sets at least the following identifiers: asset identifier, activity identifier, server identifier, user group identifier, and terminal platform identifier. The asset identifier uniquely identifies advertising materials such as character illustrations, skill videos, scene images, background music, advertising copy, and landing page components. The activity identifier identifies the game version event, holiday event, cross-server battle event, new dungeon event, or new class launch event associated with the asset. The server identifier identifies the server, battle zone, or region to which the asset applies. The user group identifier identifies the target user type, such as new users, returning users, highly active users, paying users, users who prefer competitive gameplay, or users who prefer social gameplay. The terminal platform identifier identifies the display terminal or advertising platform to which the asset applies, such as mobile, PC, console, short video platform, news feed platform, or app store ad slot.

[0043] After establishing unified feature specifications, the system acquires multimodal data of game content, real-time operational data, user feedback data, and terminal environment data. The multimodal data of game content may include character appearance, class settings, skill effects, scene atmosphere, event themes, background music, sound effect clips, and advertising copy. The real-time operational data may include the current game version, event status, remaining event time, cross-server battle status, new dungeon opening status, new class release status, server load status, and budget consumption status. The user feedback data may include exposure, clicks, viewing time, conversion, payment, comment sentiment, bounce rate, blocking, complaints, negative feedback, and material fatigue data. The terminal environment data may include terminal type, operating system, screen ratio, resolution, network status, device performance level, ad slot size, material format requirements, and platform review rules.

[0044] After acquiring the aforementioned data, the system associates and binds it based on material identifiers, activity identifiers, server identifiers, user group identifiers, and terminal platform identifiers to form material data records. Specifically, for the same advertising material, the system binds its material resource, associated activity, applicable server, target user group, and applicable terminal platform, ensuring that each material data record simultaneously reflects the material's content attributes, operational attributes, user attributes, and terminal attributes. For example, a "new profession skill demonstration video" can be bound to the video resource itself through its material identifier, to the "new profession launch" activity through its activity identifier, to the server range where the profession is available through its server identifier, to user groups that prefer melee DPS professions or high-combat-intensity gameplay through its user group identifier, and to a vertical short video platform or mobile in-feed ad slot through its terminal platform identifier. This allows the system to avoid semantic inconsistencies in materials caused by scattered multi-source data.

[0045] Furthermore, the system performs data preprocessing and feature normalization on the material data records to obtain standardized material features. The data preprocessing may include data cleaning, missing value handling, duplicate data merging, abnormal data removal, material validity verification, and time status verification. Specifically, the system can remove records with inaccessible resource addresses, invalid review status, expired activities, or mismatched terminal specifications; for data repeatedly reported by the same material on multiple platforms, the system merges them according to the material identifier and terminal platform identifier; for missing activity status, user group tags, or terminal specification information, the system can supplement them based on material metadata, historical delivery records, or platform rules; records that cannot be supplemented and affect delivery judgment are marked as unavailable or entered into a manual review queue.

[0046] The feature normalization process enables features of different types, scales, and sources to participate in subsequent unified encoding and model calculation. Specifically, the system converts features such as visual, audio, text, operational, user feedback, and terminal environment features into standard expressions within a preset range or category. For example, continuous features such as screen brightness, skill effect intensity, music rhythm intensity, click-through rate, conversion rate, negative feedback rate, server load, and network conditions are mapped to a unified scale; discrete features such as character profession, scene type, event theme, platform type, and ad placement style are converted into standard category labels; and copywriting selling points, comment sentiment, and user interest descriptions are converted into standard semantic labels. Through the above processing, data from different sources can be compared, combined, and subsequently quantum-encoded under the same feature specification.

[0047] Finally, the system organizes standardized material features into an advertising material feature set according to a unified feature specification. Specifically, the system uses material data records as the basic unit, and uniformly encapsulates the material identifier, activity identifier, region / server identifier, user group identifier, terminal platform identifier, and corresponding visual features, audio features, text semantic features, operational status features, user feedback features, and terminal adaptation features in each record to form an advertising material feature set. This advertising material feature set can be indexed by material dimension, and can also be retrieved by activity, region / server, user group, or terminal platform, thus providing a standardized data foundation for subsequent cross-modal quantum coding, dynamic quantum coding library updates, QUBO model construction, and advertising delivery scheme generation.

[0048] Through this embodiment, the system can unify the scattered multimodal content information, real-time operation information, user feedback information and terminal environment information in MMO game advertising materials into a consistent, semantically clear and computable advertising material feature set, thereby solving the problems of difficulty in associating multi-source data, inconsistent feature definitions and incomplete basic data for material search in existing methods.

[0049] In some embodiments, in step S102 above, the step of encoding visual content features into a main quantum state based on the advertising material feature set, and encoding audio features, text semantic features, and user preference features into auxiliary quantum states coupled to the main quantum state to form a cross-modal quantum coding result, specifically includes: Visual content features that characterize the advertiser's visual expression are extracted from the feature set of advertising materials, and the visual content features are discretized and weighted according to the preset visual coding rules to generate the master quantum state; Audio features, text semantic features, and user preference features associated with visual content features are extracted from the feature set of advertising materials, and encoded into auxiliary quantum states according to a unified scale conversion rule; Based on the matching relationship between visual content features and audio features, text semantic features and user preference features, the coupling weight between the master quantum state and the auxiliary quantum state is determined. The master quantum state, auxiliary quantum state, and coupling weights are associated and stored to form a cross-modal quantum encoding result that characterizes the cross-modal matching relationship of advertising materials.

[0050] In this embodiment, the system first extracts visual content features from the advertising material feature set to characterize the advertiser's visual expression. These visual content features may include character image, profession appearance, skill effects, scene type, screen color tone, screen brightness, composition center, action intensity, combat atmosphere, holiday decorative elements, and visual impact. Since MMO game advertisements typically use characters, professions, skills, and scenes as the first things users perceive in the content, the system processes these visual content features as the primary expressive features of the advertising material.

[0051] The system discretizes and weights visual content features according to preset visual encoding rules, generating a master quantum state. Specifically, the system can pre-set visual encoding rules to classify different visual content into standardized visual categories, such as "melee character classes," "ranged magic classes," "high-burst skill effects," "dark dungeon scenes," "festival scenes," and "multiplayer team battles." Simultaneously, the system assigns weights to each visual category based on the salience, screen proportion, clarity, motion intensity, and historical performance of the corresponding visual elements in the material. For continuous visual features, such as brightness, color saturation, effect intensity, or motion amplitude, the system discretizes them according to preset intervals; for categorical visual features, such as character class, scene theme, or skill attributes, the system labels them according to a unified category. After the above processing, the visual content features are converted into a computable and comparable master quantum state, used to represent the core visual expression of the advertising material.

[0052] Subsequently, the system extracts audio features, textual semantic features, and user preference features associated with visual content features from the advertising material feature set, and encodes them into auxiliary quantum states according to a unified scale conversion rule. The audio features may include background music rhythm, volume intensity, mood type, timbre style, combat sound effect intensity, and holiday sound effect elements; the textual semantic features may include semantic information such as advertising titles, advertising text, activity descriptions, welfare selling points, career selling points, combat selling points, social selling points, growth selling points, and limited-time recall copy; the user preference features may include the user's level of interest in career type, combat gameplay, social gameplay, dungeon challenges, cross-server competition, welfare activities, character appearance, and music style, as well as user feedback information such as historical clicks, viewing time, conversions, payments, bounces, and negative feedback.

[0053] When encoding auxiliary features, the system employs a unified scaling rule to enable features from different sources and with different scales, such as audio, text semantics, and user preferences, to establish comparable correlations with the master quantum state. Specifically, the system converts continuous features such as music rhythm, emotional intensity, semantic relevance, user interest intensity, and historical feedback performance into a unified intensity level; it converts discrete features such as audio style, copywriting selling points, gameplay type, and user group type into standard labels; and it marks positive and negative feedback as factors that enhance or weaken the expression of the auxiliary quantum state, respectively. Through this processing, audio features, text semantic features, and user preference features each form an auxiliary quantum state corresponding to the master quantum state.

[0054] Then, the system determines the coupling weight between the primary quantum state and the auxiliary quantum state based on the matching relationship between visual content features, audio features, text semantic features, and user preference features. Specifically, the system can judge the matching relationship from four aspects: content consistency, emotional consistency, selling point consistency, and user interest consistency. For example, when the visual content is a "cross-server team battle scene," if the background music is high-tempo battle music, the copy highlights selling points such as "ten thousand people on the same screen" and "cross-server domination," and the target user group has a high interest in competitive gameplay, then the system increases the coupling weight between the primary quantum state and the corresponding auxiliary quantum state. When the visual content is a "festival celebration scene," if the audio contains a festive atmosphere, the copy highlights limited-time benefits or festival activities, and the target user group has a high interest in benefit activities or cosmetic collection, then the system increases its coupling weight. Conversely, when skill battle scenes are accompanied by soothing music, or when the copy emphasizes social leisure but the visual content mainly presents high-intensity battles, the system decreases the corresponding coupling weight.

[0055] To ensure the usability of coupling weights, the system can also adjust the matching relationship by combining historical campaign performance and user feedback. For example, for cross-modal combinations with historically high click-through rates, conversion rates, and view completion rates, and low negative feedback, the system increases their coupling weights; for combinations with high bounce rates, blocking rates, or complaint rates, the system decreases their coupling weights. Thus, coupling weights not only reflect the surface matching relationship between content materials but also the actual acceptance level of the target users for the cross-modal combination.

[0056] Finally, the system associates and stores the master quantum state, auxiliary quantum states, and coupling weights to form a cross-modal quantum encoding result that characterizes the cross-modal matching relationship of advertising creatives. Specifically, the system uses the creative identifier as an index to bind and store the master quantum state, audio auxiliary quantum state, text semantic auxiliary quantum state, user preference auxiliary quantum state, and the coupling weights between each auxiliary quantum state and the master quantum state corresponding to the creative. Simultaneously, the system retains the activity identifier, user group identifier, terminal platform identifier, and timestamp information to update the encoding result when the operational status changes subsequently.

[0057] Through this embodiment, the system can uniformly convert the visual, audio, textual semantics, and user preferences in MMO game advertising materials into cross-modal quantum coding results with primary and secondary relationships and coupling relationships. This allows advertising materials to no longer be searched based on a single tag or a single historical indicator, but to comprehensively reflect the visual appeal, auditory emotion, textual selling points, and user interest matching degree of the materials, providing a foundation for subsequent dynamic quantum coding library generation and QUBO combination optimization.

[0058] In some embodiments, step S103 above, which involves dynamically updating the encoding primitives, coupling weights, and constraint labels in the cross-modal quantum coding results based on changes in game version, activity theme, and operational status to generate a dynamic quantum coding library, specifically includes: Obtain game version information, event configuration information, and operation monitoring information, and identify external change events based on the game version information, event configuration information, and operation monitoring information; The affected material range is determined based on external change events, and the master quantum state, auxiliary quantum state and coupling weight corresponding to the material range are retrieved from the cross-modal quantum coding results. Based on the type of change in external events, the encoding primitives corresponding to the master quantum state and auxiliary quantum state are added, replaced, or deactivated, and the coupling weights are adjusted. Based on the conditions of external change events, configure corresponding constraint labels for the cross-modal quantum encoding results; The updated encoding primitives, coupling weights, and constraint labels are indexed and stored to generate a dynamic quantum coding library.

[0059] In this embodiment, the system first acquires game version information, activity configuration information, and operational monitoring information, and identifies external change events based on this information. The game version information may include the current version number, update package content, new class opening information, new dungeon opening information, character balance adjustment information, and resource hot-update information. The activity configuration information may include festival activities, limited-time welfare activities, cross-server battle activities, returnee recall activities, new server activities, and the start time, end time, participation conditions, and reward content of these activities. The operational monitoring information may include server load, server opening status, advertising budget consumption, material review status, ad delivery consumption rate, user negative feedback, material fatigue, and platform rule changes. Based on the above information, the system determines whether an external change event has occurred, such as a new version launch, a new class release, the start of a festival activity, the start of a cross-server battle, high load on a certain server, expired material review, or an expired activity.

[0060] After identifying external change events, the system determines the scope of affected materials based on these events and retrieves the corresponding master quantum state, auxiliary quantum state, and coupling weights from the cross-modal quantum coding results. Specifically, the system can determine the affected objects based on event identifiers, material identifiers, server identifiers, user group identifiers, and terminal platform identifiers. For example, when the external change event is "new profession launched," the affected material scope may include the character illustrations, skill videos, profession introduction text, trial landing pages, and user groups with similar preferences for the new profession; when the external change event is "Spring Festival event begins," the affected material scope may include Spring Festival-themed scenes, holiday music, limited-time benefit text, and holiday event landing pages; when the external change event is "high server load in a certain region," the affected material scope may include advertising materials pointing to that region or related event entrances. Based on this, the system retrieves the corresponding master quantum state, audio auxiliary quantum state, text semantic auxiliary quantum state, user preference auxiliary quantum state, and their coupling weights from the existing cross-modal quantum coding results.

[0061] Subsequently, based on the type of change in external events, the system adds, replaces, or deactivates the encoding primitives corresponding to the primary and auxiliary quantum states, and adjusts the coupling weights. These encoding primitives can be understood as the basic feature units constituting the quantum encoding result, such as character visual primitives, profession tag primitives, skill effect primitives, scene atmosphere primitives, music emotion primitives, copywriting selling point primitives, and user preference primitives. When new content is launched, the system adds encoding primitives corresponding to the new content, such as adding encoding primitives for "new profession name," "new profession skill effect," "new dungeon scene," and "limited-time mount appearance." When resource replacement or version adjustment events occur, the system replaces the old version's character image, old skill performance, or old event copywriting with the corresponding primitives for the new version. When events expire, materials are removed, or review fails, the system deactivates the relevant encoding primitives, preventing them from participating in subsequent material search and combination optimization.

[0062] When adjusting coupling weights, the system enhances or weakens the impact of external events on the matching relationships of content. For example, during the launch of a new profession, the system enhances the coupling weight between the new profession's visual primitives and the profession's introduction text, battle music, trial button, and related user preference primitives; during holiday events, the system enhances the coupling weight between the holiday scene's visual primitives and the holiday music, limited-time benefit text, and returning user preference primitives; when a piece of content experiences high negative feedback or increased fatigue, the system reduces the coupling weight between the visual primitives of that piece of content and the corresponding text, audio, and user preference primitives. In this way, the system enables the quantum encoding results to reflect the current delivery value in real time as the game's operational status changes.

[0063] Furthermore, the system configures corresponding constraint tags for cross-modal quantum encoding results based on the delivery conditions of external change events. These constraint tags indicate the usage conditions of a particular encoding result in subsequent QUBO model construction and material combination search. Constraint tags may include "Activity Validity Tag," "Region / Server Available Tag," "Region / Server Traffic Limiting Tag," "Platform Compatibility Tag," "Approval Tag," "Budget Limitation Tag," "Frequency Limitation Tag," "Material Fatigue Tag," "Negative Feedback Limitation Tag," and "Forced Priority Tag." For example, for materials available only during the activity period, the system configures an activity validity period tag; for materials only applicable to mobile vertical ad slots, the system configures a terminal platform compatibility tag; for regions with high server load, the system configures a traffic limiting tag; for materials that have failed approval or whose activity has ended, the system configures a disabled or deactivated tag; and for materials on new professions that are being heavily promoted by the official platform, the system can configure a priority delivery tag.

[0064] Finally, the system indexes and stores the updated encoding primitives, coupling weights, and constraint labels, generating a dynamic quantum coding library. Specifically, the system uses material identifiers, activity identifiers, server identifiers, user group identifiers, terminal platform identifiers, and update times as indexes to associate and save the updated master quantum state, auxiliary quantum state, coupling weights, and constraint labels. The dynamic quantum coding library supports rapid retrieval by activity, version, server, user group, and terminal platform, and also supports directly reading the currently valid encoding results and constraints during subsequent QUBO model construction.

[0065] Through this embodiment, the system can promptly reflect game version updates, event theme changes, and operational status changes in the cross-modal quantum encoding results, avoiding reliance on outdated materials or static tags for ad creative searches. Simultaneously, the dynamic quantum encoding library provides a real-time, effective, and operationally constrained data foundation for subsequent ad creative combination optimization, thereby improving the timeliness and adaptability of MMO game ad creative searches.

[0066] In some embodiments, step S104 above, which involves constructing a QUBO model based on a dynamic quantum coding library, with advertising effectiveness and experience consistency as optimization objectives and operational capacity and platform adaptation as dynamic constraints, specifically includes: Based on the activity identifier, target user group, delivery channel, terminal type, and delivery time window of the current delivery task, the active advertising creative code records are selected from the dynamic quantum coding library, and the selected advertising creative code records are mapped as creative candidate variables; Historical performance data, campaign matching data, and cross-modal consistency data are extracted from advertising creative coding records to generate performance evaluation weights and experience consistency weights. Extract operational constraint parameters and platform adaptation constraint parameters corresponding to server capacity, region / server status, delivery frequency, platform review, ad slot size, and material format from the constraint tags of the dynamic quantum coding library; Target items for prioritizing high-quality materials are set based on performance evaluation weights and experience consistency weights, and penalty items for limiting incompatible materials are set based on operational constraint parameters and platform adaptation constraint parameters. The QUBO model combines candidate variables, target terms, penalty terms, and collaborative or conflicting relationships between materials.

[0067] In this embodiment, the system first receives the current campaign information. This campaign information may include an activity identifier, target user group, campaign channel, terminal type, and campaign time window. For example, the campaign may be a "New Profession Launch Event," a "Returning User Recall Event," a "Spring Festival Welfare Event," or a "Cross-Server Battle Pre-launch Event"; the target user group may be new users, returning users, highly active users, paying users, users who prefer competitive gameplay, or users who prefer social gameplay; the campaign channel may be a short video platform, a news feed advertising platform, an app store, a community platform, or in-game / out-of-game integrated channels; the terminal type may be a mobile device, a PC, or a tablet; and the campaign time window is used to limit the start and end times during which the materials can be used for campaign placement.

[0068] Based on the aforementioned campaign information, the system filters active ad creative codes from the dynamic quantum coding library. Specifically, the system can filter creatives related to the current campaign based on campaign identifiers, creatives matching user preferences based on target user groups, and creatives meeting ad placement specifications, creative formats, and platform review requirements based on campaign channels and terminal types. It also excludes creatives whose campaigns have expired, are not yet active, or are inactive based on the campaign time window. Furthermore, for creatives with constraint tags such as "approved," "not applicable to certain regions / servers," "creative fatigue," "negative feedback restriction," or "terminal incompatibility," the system can exclude them or set higher penalties in subsequent models. The filtered ad creative codes are mapped to candidate creative variables, each indicating whether the ad creative is selected for the current campaign.

[0069] Subsequently, the system extracts historical performance data, campaign matching data, and cross-modal consistency data from the advertising creative coding records to generate performance evaluation weights and experience consistency weights. The historical performance data may include impressions, clicks, click-through rate, view completion rate, conversion rate, registration rate, paid conversion rate, retention performance, negative feedback rate, and creative fatigue. The campaign matching data may include the degree of matching between the role, profession, scene, benefits, and gameplay selling points expressed by the creative and the current campaign theme. The cross-modal consistency data may include the coupling weights between the main visual quantum state and the auxiliary audio quantum state, text semantic auxiliary quantum state, and user preference auxiliary quantum state. Based on the above data, the system sets performance evaluation weights and experience consistency weights for each creative candidate variable. Performance evaluation weights are used to characterize the click, conversion, or retention benefits that the creative may bring; experience consistency weights are used to characterize whether the visuals, music, copy, and user interests of the creative are consistent, avoiding issues such as "the visuals depict intense battles but the copy emphasizes casual social interaction" or "the promotional creative matches the music of competitive battles."

[0070] Furthermore, the system extracts operational constraint parameters and platform adaptation constraint parameters corresponding to server load, region / server status, delivery frequency, platform review, ad slot size, and creative format from the constraint tags in the dynamic quantum coding library. The operational constraint parameters may include server load thresholds, region / server open status, campaign validity period, budget consumption status, creative delivery frequency, user reach frequency, creative fatigue, and negative feedback restrictions. The platform adaptation constraint parameters may include ad slot size, landscape / portrait screen requirements, video duration requirements, image aspect ratio, file size, encoding format, landing page requirements, platform review status, and channel prohibition rules. For example, when a region / server's server load is too high, ad creatives strongly associated with that region / server can be configured with rate limiting constraints; when a creative has reached the platform's frequency limit, that creative is restricted from further selection in the current delivery window; when the creative format does not conform to the ad slot specifications of the delivery channel, the system considers it an incompatible creative.

[0071] After obtaining the performance evaluation weight, experience consistency weight, operational constraint parameters, and platform adaptation constraint parameters, the system sets target items to prioritize the selection of high-quality materials and penalty items to limit incompatible materials. The target items are used to encourage the model to prioritize materials with good historical performance, high activity theme matching, strong cross-modal consistency, and alignment with target user preferences. For example, for a "New Profession Launch Event," if a skill demonstration video has a high completion rate and its visual content, battle music, and profession selling point copy are highly consistent with target user preferences, then this material will receive a high selection priority in the target items. The penalty items are used to reduce or exclude materials that do not meet operational constraints and platform adaptation requirements. For example, for materials that have failed review, expired events, are unavailable in certain regions / servers, have incompatible terminal specifications, excessive negative feedback, excessive material fatigue, or exceed frequency limits, the system sets higher penalties in the penalty items to make them less likely to be selected by the model; for materials with only minor mismatches, such as those requiring cropping or with frequencies close to the limit, lower penalties can be set, allowing the model to make trade-offs when better materials are unavailable.

[0072] Furthermore, the system constructs relationships between variables based on the synergistic or conflicting relationships between creative materials. Synergistic relationships refer to the ability of multiple creative materials appearing simultaneously to improve overall campaign performance or user experience consistency. For example, the main visual video, promotional copy, and landing page components within the same campaign may have a consistent theme, or the same user group may show high interest in promotional materials and corresponding strategy copy for a particular profession. Conflicting relationships refer to the potential for duplicate reach, theme conflicts, platform limitations, or a decline in user experience when multiple creative materials are selected simultaneously. For example, the same user may be repeatedly shown the same character's creative material within a short period, or the same ad group may contain creative materials with different campaign themes, different promotional offers, or mutually exclusive server entry points. The system incorporates these synergistic or conflicting relationships into the QUBO model, ensuring that the model considers not only the quality of individual creative materials but also the overall performance of the combined materials.

[0073] Finally, the system combines candidate creative variables, target items, penalty items, and synergistic or conflicting relationships between creatives into a QUBO model. This QUBO model is used to search the candidate creative set for creative combinations that meet current operational constraints and platform adaptation requirements, while maintaining good consistency in ad performance and user experience. Through this model, the system can avoid the problem of selecting creatives solely based on click-through rates or human experience, and can quickly reconstruct an optimized model adapted to the current delivery task based on a dynamic quantum coding library, even when game versions, event states, server capacity, and platform rules change.

[0074] Through this embodiment, the system unifies the multimodal material representation, historical deployment effects, user preferences, operational carrying capacity status, and platform adaptation conditions in the dynamic quantum coding library into candidate variables, optimization objectives, and constraint penalties in the QUBO model, thereby providing a clear and executable model foundation for subsequent quantum annealing, simulated annealing, or other combined optimization solutions.

[0075] Furthermore, the step of extracting historical performance data, campaign matching data, and cross-modal consistency data from advertising creative coding records to generate performance evaluation weights and experience consistency weights specifically includes: Extract historical performance data corresponding to exposure, clicks, conversions, retention, paid subscriptions, recalls, negative feedback, and creative fatigue from the historical delivery logs associated with the advertising creative coding records. Perform anomaly filtering and time decay processing on the historical performance data to obtain a basic advertising performance score. Extract activity matching data related to the current event theme, character profession, gameplay selling points, benefits, and target user group from the advertising creative coding records, and compare the activity matching data with the event configuration information of the current delivery task to adjust the basic score of advertising performance; Consistency data between visual content, audio sentiment, text semantics, and user preferences is extracted from cross-modal quantum coding results in advertising creative coding records, and an experience consistency score is generated based on the consistency data. The performance evaluation weights are generated based on the adjusted basic advertising performance score, and the experience consistency weights are generated through the experience consistency score.

[0076] In this embodiment, the system first extracts historical performance data from the historical delivery logs associated with the advertising creative's encoding record. The historical delivery logs can include delivery records of the advertising creative across different channels, user groups, terminal types, and delivery time windows. The historical performance data can include metrics such as impressions, clicks, click-through rate (CTR), conversions, registrations, next-day retention, subsequent retention, number of paid subscriptions, payment amount, number of successful user reactivation attempts, number of negative feedbacks, number of blocking attempts, number of complaints, and creative fatigue. Creative fatigue can be determined based on the number of times the same creative is repeatedly exposed within a certain period, the rate of decline in CTR, the rate of decline in view completion rate, and the increase in negative feedback, and is used to characterize the degree to which users' interest in the creative diminishes.

[0077] To prevent abnormal data from affecting subsequent optimization, the system performs anomaly filtering and time decay processing on historical performance data. Anomaly filtering can include removing invalid impressions, abnormal clicks, fraudulent traffic, abnormally high conversion rates in a short period, missing statistical data, and unstable data due to insufficient sample size. For example, when a creative generates an abnormally high click-through rate with extremely low impressions, the system can reduce its credibility; when an abnormal click fluctuation occurs on a channel, the system can exclude data from the corresponding time period. Time decay processing assigns higher reference value to recent campaign data and lowers the reference value of earlier data to adapt to changes in game version, user interests, and platform traffic environment. After processing, the system obtains a basic ad performance score, which is used to represent the overall performance of the creative in historical campaigns.

[0078] Subsequently, the system extracts activity matching data related to the current event theme, character class, gameplay highlights, benefits, and target user group from the advertising creative coding records, and compares this activity matching data with the event configuration information of the current campaign. The activity matching data may include event tags, character class tags, scene tags, gameplay tags, benefits tags, copywriting highlight tags, and applicable user group tags corresponding to the creative. The event configuration information of the current campaign may include the current event name, event theme, featured class, featured gameplay, reward content, event duration, target user type, and advertising channel.

[0079] During the comparison process, the system determines whether the content expressed by the ad creative is consistent with the current campaign objective. For example, if the current campaign objective is a "New Class Launch Event," and the visual content of the creative includes the new class character and skill effects, the text semantics highlight the class's characteristics, and the target audience is users interested in similar classes or combat gameplay, then the creative has a high degree of relevance to the current campaign, and the system will improve the basic score for ad performance. Conversely, if the current campaign objective is a "Spring Festival Welfare Event," but the creative mainly focuses on cross-server competitive combat and lacks festive elements and welfare copy, then the system will lower the creative's score for the current campaign objective. Through this adjustment, the system can avoid selecting creatives solely based on historically high click-through rates while ignoring the suitability of the current campaign theme.

[0080] Next, the system extracts consistency data between visual content, audio sentiment, text semantics, and user preferences from the cross-modal quantum coding results in the advertising material encoding record, and generates an experience consistency score based on this consistency data. The consistency data can originate from the coupling weights between the main quantum state and each auxiliary quantum state, and their corresponding matching tags. Specifically, the system analyzes whether the visual content is consistent with the audio sentiment, whether the visual content is consistent with the text semantics, whether the text selling points are consistent with user preferences, and whether the overall material expression aligns with the target user's interest in MMO game content. For example, if the visual content is an intense team battle scene, the audio is high-paced battle music, the text semantics emphasize "cross-server battles" and "multiplayer on the same screen," and the target user prefers competitive gameplay, then the material has high experience consistency; if the visual content is a dark dungeon scene, but the audio is light holiday music, and the copy emphasizes social leisure, then the experience consistency is low.

[0081] When generating experience consistency scores, the system can also incorporate user negative feedback and viewing behavior for adjustments. For example, if a cross-modal combination, despite consistent content tags, exhibits high bounce rates, blocking rates, or complaint rates in historical campaigns, the system can lower its experience consistency score. Conversely, if users demonstrate high completion rates and conversion rates for a particular combination of visuals, audio, and text, the system can raise its experience consistency score. Therefore, the experience consistency score reflects both the coordination of the various modal expressions within the content and the user's actual acceptance of the combined experience.

[0082] Finally, the system generates performance evaluation weights based on the adjusted basic ad performance score and experience consistency weights based on the experience consistency score. Performance evaluation weights are used to prioritize ad performance within the target items of the QUBO model, making it easier to select ad creatives with good historical performance and a good match for the current campaign. Experience consistency weights are used to reflect the degree of experience coordination of ad creatives within the target items or ad creative combinations of the QUBO model, making it easier to select ad creative combinations where visuals, audio, copy, and user interests are consistent. For ad creatives with high performance but poor experience consistency, the system can lower their overall priority in the model; for new ad creatives with average historical performance but a high match for the current campaign and strong experience consistency, the system can provide appropriate selection opportunities to support the cold start of new versions or campaigns.

[0083] Through this embodiment, the system can comprehensively transform historical campaign performance, current campaign adaptation, and cross-modal experience consistency into weight parameters that can be used by the QUBO model, so that subsequent material search not only pursues click or conversion effects, but also takes into account the consistency between ad expression and user experience.

[0084] Furthermore, the combination of candidate variables, target terms, penalty terms, and collaborative or conflicting relationships between materials into a QUBO model specifically includes: Generate synergistic relationships based on the complementary relationships between materials in terms of roles, gameplay, benefits, channels, and conversion paths; Conflict relationships are generated based on the incompatibilities between materials in terms of copywriting, visuals, event versions, target audiences, and platform rules; Candidate variables, target terms, penalty terms, synergistic relationships, and conflict relationships are written into a quadratic unconstrained binary optimization structure to form a QUBO model for solving the target ad creative combination.

[0085] In this embodiment, after determining candidate variables for creative materials, setting target terms, and setting penalty terms, the system further analyzes the combination relationships between different advertising creative materials. Each candidate variable for creative materials corresponds to an advertising creative material code record, used to indicate whether the material is selected into the current advertising campaign. The target terms are used to increase the probability of high-quality creative materials being selected, and the penalty terms are used to restrict the selection of creative materials that do not meet operational constraints or platform adaptation requirements. Based on this, the system introduces collaborative and conflict relationships between creative materials so that the model can evaluate the overall effect of the creative material combination, rather than just the effect of individual creative materials.

[0086] First, the system generates synergistic relationships based on the complementary relationships between materials in terms of roles, gameplay, benefits, channels, and conversion paths. The complementary relationship refers to the mutually reinforcing relationship that occurs when two or more materials are deployed simultaneously, in terms of content expression, user perception, or conversion path. For example, in a "New Profession Launch Event," one material highlights the new profession's character image, another showcases the profession's high-burst skill effects, and a third emphasizes "receive a profession-exclusive gift pack upon login" through copywriting. These materials complement each other in terms of character presentation, gameplay experience, and benefit incentives, and can be identified as having a synergistic relationship. Similarly, a short video material used to attract user interest, an in-feed image material used to reinforce benefit recall, and a landing page material used to complete reservation or download conversions—these three are interconnected in the conversion path and can also be identified as having a synergistic relationship.

[0087] When generating collaborative relationships, the system can make judgments based on character tags, gameplay tags, benefit tags, channel tags, landing page tags, and user group tags recorded in the dynamic quantum coding library. If two materials have the same theme but different focuses, such as one emphasizing "cross-server team battles" and the other emphasizing "rewards for thousands of people on the same screen," they can be marked as collaborative. If one material is suitable for short video channels to stimulate interest, and another material is suitable for search or information flow channels to facilitate conversion, and both point to the same activity or the same conversion path, they can also be marked as collaborative. For materials with collaborative relationships, the system sets up relationships in the QUBO model that encourage their co-selection, making the solution results more inclined to form a complete and coherent combination of advertising materials.

[0088] Secondly, the system generates conflict relationships based on the incompatibilities between materials in terms of copywriting, visuals, event versions, target audiences, and platform rules. These incompatibilities refer to situations where two or more materials are simultaneously selected for the same campaign, potentially leading to informational contradictions, user interference, misleading campaigns, resource waste, or platform violations. For example, if one material's copy states "Limited-time offer for three days," and another states "Long-term benefits available," and both refer to the same event but have different benefit descriptions, there is a copywriting conflict. Similarly, if one material displays an older version of a character model, and another displays a newer version, and the current version has been updated, there is a version conflict between the old and new versions. Furthermore, repeatedly reaching the same user group with multiple highly similar materials within a short period can lead to excessive frequency and user fatigue, and can also be identified as a target audience conflict.

[0089] Furthermore, platform rule conflicts also need to be incorporated into the model. For example, a certain advertising channel may prohibit the use of specific exaggerated expressions, or an ad placement may require that the creative materials not contain excessively small text, click-through buttons, or content that does not meet size specifications. If a combination of creative materials violates platform rules when placed on the same channel, the system will mark it as a conflict. As another example, different campaign versions may be mutually exclusive. When the current campaign is a "Spring Festival Welfare Campaign," expired "Anniversary Celebration Welfare Campaign" creative materials should not appear together with the current campaign's creative materials to avoid user misunderstanding. For creative materials with conflicting relationships, the system sets a relationship in the QUBO model to inhibit their co-selection, ensuring that the solution results avoid inconsistent or unsuitable creative material combinations.

[0090] Subsequently, the system incorporates candidate material variables, target terms, penalty terms, collaborative relationships, and conflict relationships into a quadratic unconstrained binary optimization structure, forming a QUBO model for solving the target ad creative combination. Specifically, candidate material variables constitute the set of binary variables in the model, with each variable representing the selection state of the corresponding material; the target term reflects the advertising effectiveness weight and experience consistency weight of a single material; the penalty term reflects the degree to which a single material or combination of materials violates constraints such as operational capacity, delivery frequency, review status, ad placement size, material format, and region / server status; collaborative relationships are used to increase the selection priority of material combinations with association enhancement effects; and conflict relationships are used to reduce or eliminate the possibility of mutually exclusive materials appearing simultaneously.

[0091] In practical applications, the system can uniformly store the above relationships in the variable terms and inter-variable relationship terms of the QUBO model. For the priority or limitations of a single material itself, it is written into the single-variable term corresponding to the candidate variable of that material; for the synergy or conflict between two materials, it is written into the two-variable relationship term corresponding to the candidate variables of the two materials. In this way, the model can simultaneously consider "which materials are better individually" and "which combinations of materials are more suitable" when solving the problem. For example, for a new career activity, the model might prioritize a group of synergistic materials: "career showcase video," "skill demonstration short video," "career gift pack copywriting image," and "career appointment landing page," while excluding old career promotional images, expired benefit images, or video materials incompatible with platform rules.

[0092] In this embodiment, the QUBO model can uniformly express the individual value of advertising creatives, operational constraints, platform compatibility conditions, and the combination relationships between creatives as a solvable quadratic unconstrained binary optimization structure. Therefore, subsequent solutions using quantum annealing, simulated annealing, or other combinatorial optimization methods can obtain the target advertising creative combination from the candidate creative set, achieving a better balance between advertising effectiveness, user experience consistency, operational feasibility, and platform compliance in the final campaign.

[0093] In some embodiments, step S105 above, which involves solving the QUBO model using quantum annealing or a quantum heuristic algorithm to obtain the target advertising material combination, specifically includes: Read the candidate variables, target terms, penalty terms, and cooperative or conflict relationships between materials in the QUBO model, and determine the quantum annealing solution mode or quantum heuristic solution mode based on available computing resources; When the quantum annealing solution mode is determined, the QUBO model is converted into an input structure that can be recognized by the quantum annealing device, and the annealing rounds, sampling times and annealing time are set to obtain multiple candidate solutions; When the quantum heuristic solution mode is determined, multiple material combination states are initialized based on the QUBO model, and the material combination states are updated through iterative search to obtain multiple candidate solutions; Multiple candidate solutions are constrained, validated, and ranked based on advertising effectiveness, user experience consistency, operational capacity, and platform compatibility to obtain the ranking result. The candidate solutions that meet the preset conditions in the sorting results are restored to a set of advertising material identifiers, and the target advertising material combination is generated.

[0094] In this embodiment, the system first reads the constructed QUBO model. The QUBO model includes candidate variables for advertising materials, target terms, penalty terms, and synergistic and conflicting relationships between materials. Specifically, candidate variables indicate whether an advertising material is selected; target terms reflect the consistency of advertising effectiveness and user experience; penalty terms reflect constraints such as operational capacity, server / region status, frequency of placement, platform review, ad placement size, and material format; and synergistic and conflicting relationships reflect the enhancement or repulsion effects of combining materials. After reading the above information, the system determines whether to use quantum annealing or quantum heuristic solution mode based on available computing resources.

[0095] Specifically, when the system can access a quantum annealing device, and the size of candidate variables, model sparsity, and device availability meet the requirements, the quantum annealing solution mode can be adopted. When the quantum annealing device is unavailable, the size of candidate variables exceeds the direct capacity of the device, a fast solution needs to be obtained on a local server or general-purpose cloud computing resources, or multiple rounds of online real-time optimization are required, the quantum heuristic solution mode can be adopted. The quantum heuristic solution mode may include simulated annealing, quantum heuristic annealing, tabu search, genetic search, or other combined optimization search methods suitable for QUBO models.

[0096] When the quantum annealing solution mode is determined, the system converts the QUBO model into an input structure recognizable by the quantum annealing device. This input structure includes at least binary variable information corresponding to the candidate creative variables, variable weight information corresponding to the target and penalty terms, and inter-variable coupling information corresponding to cooperative and conflict relationships. If the quantum annealing device has limitations on the number of variables or connections, the system can perform embedding or block processing on the QUBO model to map the model to a topology supported by the quantum annealing device. Then, the system sets the annealing rounds, sampling times, and annealing time. The annealing rounds control the number of times the overall solution is repeated, the sampling times acquire multiple possible creative combinations, and the annealing time controls the evolution duration of each annealing process. Through multiple rounds of sampling, the system obtains multiple candidate solutions, each corresponding to a set of selected advertising creative variables.

[0097] When the quantum heuristic solution mode is determined, the system initializes multiple creative combination states based on the QUBO model. These creative combination states can be generated randomly, or based on historically high-quality campaign strategies, creatives strongly relevant to the current campaign, manually pre-selected creatives, or the results of the previous optimization round. Each creative combination state represents a set of initial ad creative selection results. Subsequently, the system updates the creative combination states through iterative search. During the update process, the system can flip the selection state of a single creative, or simultaneously adjust a group of creatives with synergistic or conflicting relationships, and determine whether the adjusted combination is better based on the objective term, penalty term, and relationships between variables in the QUBO model. For combinations that are locally poor but may escape the local optimum, the system can retain a certain probability according to a preset search strategy to improve global optimization capabilities. After multiple iterations, the system obtains multiple candidate solutions.

[0098] After obtaining multiple candidate solutions, the system performs constraint verification and sorting. Constraint verification is used to determine whether the candidate solutions meet the hard requirements of the current campaign, including but not limited to platform approval, ad format conforming to ad placement requirements, size ratio conforming to channel specifications, campaign time within the valid window, server status being advisable, server capacity not exceeding thresholds, user reach frequency not exceeding limits, and consistency with campaign version. Candidate solutions that violate the hard requirements can be directly eliminated by the system; for candidate solutions with minor deviations that can be resolved by reducing frequency, replacing ad placements, or post-processing adaptation, the system can retain them but lower their sorting priority.

[0099] During the ranking process, the system comprehensively considers ad performance, user experience consistency, operational capacity requirements, and platform compatibility. Ad performance is determined based on the performance evaluation weights of the creatives in the candidate solutions and their combined gains; user experience consistency is determined based on the overall coordination of visual content, audio sentiment, text semantics, and user preferences; operational capacity requirements are determined based on server load, server / region status, frequency control, and budget status; and platform compatibility is determined based on review results, ad placement size, creative format, channel rules, and landing page requirements. Through this comprehensive ranking, the system obtains a ranking result for multiple candidate solutions.

[0100] Finally, the system restores the candidate solutions that meet preset conditions in the ranking results into a set of advertising creative identifiers, generating a target advertising creative combination. The preset conditions may include a candidate solution ranking within a preset number, a comprehensive score reaching a threshold, all hard constraints being met, the number of creatives meeting the delivery requirements, or a candidate solution having a preset improvement over the current online delivery plan. The restoration process refers to mapping the binary variable selection state in the candidate solution back to advertising creative identifiers in a dynamic quantum coding library, such as creative ID, version number, channel-compatible version, landing page identifier, and event identifier. The resulting target advertising creative combination may include video creatives, image creatives, copywriting creatives, audio creatives, landing page components, and their corresponding delivery channel configurations.

[0101] In this embodiment, the system can flexibly choose between quantum annealing or quantum heuristics to solve the QUBO model based on computing resources. After solving, it verifies and sorts the candidate creative combinations in conjunction with advertising business constraints, thereby outputting a combination of target advertising creatives that can be actually deployed. This approach utilizes the QUBO model's ability to express complex combination relationships while ensuring that the final result meets multiple requirements such as advertising effectiveness, user experience consistency, operational capacity, and platform compatibility.

[0102] In some embodiments, step S106 above, which involves decoding and matching the target advertising material combination, and generating a corresponding advertising delivery plan according to differences in the terminal environment, specifically includes: Based on the variable mapping relationship in the dynamic quantum coding library, the value results of the candidate variables in the target advertising material combination are decoded to obtain the corresponding advertising material information, which includes advertising material identifier, material version, material type and material resource address; Based on the current campaign theme, target user group, ad placement type, delivery channel, and delivery time window, the decoded ad creative information is verified for content tags, review status, creative validity period, and resource availability to determine the matching creatives. Obtain terminal environment difference information corresponding to terminal operating system, screen ratio, resolution, network status, device performance and platform material specifications, and adapt the display ratio, material format, video bitrate, copy language and landing page path of the matching materials according to the terminal environment difference information; The adapted creative materials are then bound to the delivery channels, ad placement types, display formats, delivery time periods, and exposure frequencies to generate an advertising delivery plan for the corresponding terminal environment.

[0103] In this embodiment, after obtaining the target advertising creative combination, the system first decodes the values ​​of the candidate variables in the target advertising creative combination according to the variable mapping relationship in the dynamic quantum coding library. The variable mapping relationship is used to record the correspondence between candidate variables in the QUBO model and actual advertising creatives. For example, a candidate variable can correspond to a specific version of a video creative, or it can correspond to a specific image creative, text creative, audio creative, or landing page component. Based on the selected candidate variable, the system queries the dynamic quantum coding library to obtain the corresponding advertising creative information. The advertising creative information may include advertising creative identifier, creative version, creative type, and creative resource address, and may also include auxiliary information such as creative tags, applicable channels, applicable ad positions, review status, validity period, size ratio, language version, and landing page identifier.

[0104] Subsequently, the system matches and verifies the decoded ad creative information based on the current campaign's theme, target user group, ad placement type, campaign channel, and campaign time window to determine matchable creatives. Specifically, the system first verifies whether the creative's content tags match the current campaign theme. For example, if the current campaign is a "New Profession Launch Event," the creative should include tags such as corresponding profession, skills, benefits, or version updates; when the campaign targets returning users, the creative should include tags such as return rewards, old player recall, and version benefits. Secondly, the system verifies the creative's approval status; only creatives that have passed approval or meet the current channel's availability requirements can proceed to subsequent adaptation. Thirdly, the system verifies the creative's validity period to avoid using expired campaigns, offline characters, invalid benefits, or outdated visuals. The system also verifies resource availability, including whether the creative resource address is accessible, whether the file is complete, whether the transcoded version exists, whether the landing page can be opened, and whether resource loading meets expectations. After these verifications, the system obtains a set of matchable creatives.

[0105] After identifying matching materials, the system obtains terminal environment difference information. This terminal environment difference information may include the terminal operating system, screen ratio, resolution, network status, device performance, and platform material specifications. For example, the terminal operating system may include Android, iOS, or other systems; the screen ratio may include portrait, landscape, full-screen, foldable, etc.; resolution may be used to determine whether the display version is high-definition, standard-definition, or low-definition; network status may include Wi-Fi, 5G, 4G, or weak network environments; device performance may include high-performance devices, mid-to-low-end devices, or low-memory devices; and platform material specifications may include the requirements of different advertising platforms regarding video duration, cover size, file size, text ratio, button style, and navigation methods.

[0106] The system adapts matching materials based on differences in terminal environments. For display ratio, the system can select horizontal, vertical, or rectangular materials based on the ad placement and screen ratio, and can also perform safe area cropping to prevent key characters, promotional text, or buttons from being obscured. For material format, the system can select video, GIF, image, or a combination of text and images according to channel requirements and match the corresponding file format. For video bitrate, the system can select a high-definition, high-bitrate version on high-performance devices and in good network environments, and a low-bitrate or short-duration version on weak network or low-end devices to reduce loading waits and playback stuttering. For copy language, the system can select Simplified Chinese, Traditional Chinese, English, or other language versions based on the target region, user language preferences, or platform requirements, ensuring consistency in promotional descriptions, event times, and button text. For landing page paths, the system can select the corresponding download page, reservation page, return page, or deep link path based on the operating system and channel type. For example, Android users are redirected to the Android download address, iOS users are redirected to the app store or reservation page, and users who have already installed the app can enter a specific event page within the game via a deep link.

[0107] Finally, the system binds the adapted, matchable creatives with the delivery channel, ad type, display format, delivery time period, and exposure frequency to generate an advertising delivery plan for the corresponding terminal environment. This binding refers to determining the specific delivery content and parameters for different terminal environments and channel combinations. For example, for vertical screen feed ads on short video channels, the system can bind vertical video creatives, short copy, event landing pages, high-activity evening delivery times, and preset exposure frequencies; for search ads or feed image / text ads, the system can bind image creatives, title copy, benefit descriptions, and download landing pages; for users in weak network environments, the system can bind lightweight image creatives or low-bitrate short video creatives to ensure a high display success rate.

[0108] The generated ad placement plan can include elements such as creative identifiers, creative versions, resource addresses, placement channels, ad placement types, display formats, placement time periods, exposure frequency, target user groups, landing page paths, and device adaptation parameters. The system can also generate multiple sub-plans for different device environments, such as plans for high-performance Android devices, iOS devices, low-end devices with weak network conditions, landscape ad placements, and portrait ad placements, allowing the system to select the appropriate plan in real time based on user requests.

[0109] Through this embodiment, the system can convert the abstract candidate material variables output by the QUBO model into actual, deployable advertising materials, and complete content matching and display adaptation by combining the campaign, channel, and terminal environment. Therefore, the target advertising material combination not only performs better in the optimization model, but also meets the requirements of review compliance, resource availability, terminal adaptation, and user experience during actual deployment.

[0110] Reference Figure 2 An embodiment of the present invention provides an advertising material search system 2 based on a quantum heuristic algorithm, the system specifically comprising: The data acquisition module 201 is used to acquire multimodal data of game content, real-time operation data, user feedback data and terminal environment data of MMO games, and generate advertising material feature sets based on unified feature specifications; The quantum coding module 202 is used to encode visual content features into the main quantum state based on the feature set of advertising materials, and to encode audio features, text semantic features and user preference features into auxiliary quantum states coupled with the main quantum state, forming a cross-modal quantum coding result; The dynamic update module 203 is used to dynamically update the encoding primitives, coupling weights and constraint labels in the cross-modal quantum coding results according to changes in game version, event theme and operation status, and generate a dynamic quantum coding library. Model building module 204 is used to build a QUBO model based on a dynamic quantum coding library, with advertising effectiveness and experience consistency as optimization objectives and operational capacity and platform adaptation as dynamic constraints. The model solving module 205 is used to solve the QUBO model using quantum annealing or quantum heuristic algorithms to obtain the target advertising creative combination; The ad delivery module 206 is used to decode and match the target ad creative combination, and generate corresponding ad delivery plans according to the differences in terminal environment.

[0111] It is understandable that, such as Figure 1 The content of the advertising material search method embodiment based on the quantum heuristic algorithm shown is applicable to the advertising material search system embodiment based on the quantum heuristic algorithm. The specific functions implemented by the advertising material search system embodiment based on the quantum heuristic algorithm are the same as those shown in the example. Figure 1 The illustrated embodiment of the advertising creative search method based on quantum heuristic algorithms is the same, and the beneficial effects achieved are the same as those shown. Figure 1 The beneficial effects achieved by the illustrated embodiment of the advertising creative search method based on quantum heuristic algorithms are also the same.

[0112] It should be noted that the information interaction and execution process between the above systems are based on the same concept as the method embodiments of the present invention. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0113] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0114] Reference Figure 3 The present invention also provides a computer device 3, including: a memory 302 and a processor 301, and a computer program 303 stored in the memory 302. When the computer program 303 is executed on the processor 301, it implements the advertising material search method based on quantum heuristic algorithm as described in any of the above methods.

[0115] The computer device 3 may be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will understand that... Figure 3 The computer device 3 is merely an example and does not constitute a limitation on the computer device 3. It may include more or fewer components than shown in the figure, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0116] The processor 301 can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0117] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as a hard disk or memory of the computer device 3. In other embodiments, the memory 302 may be an external storage device of the computer device 3, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 3. Furthermore, the memory 302 may include both internal and external storage units of the computer device 3. The memory 302 is used to store the operating system, applications, boot loader, data, and other programs, such as the program code of the computer program. The memory 302 can also be used to temporarily store data that has been output or will be output.

[0118] This invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the advertising material search method based on a quantum heuristic algorithm as described in any of the above methods.

[0119] In this embodiment, if the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0120] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0121] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0122] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

Claims

1. A method for searching advertising creatives based on a quantum heuristic algorithm, characterized in that, The method specifically includes: Acquire multimodal data of game content, real-time operational data, user feedback data, and terminal environment data of MMO games, and generate advertising material feature sets based on unified feature specifications; Based on the feature set of advertising materials, visual content features are encoded into the main quantum state, and audio features, text semantic features and user preference features are encoded into auxiliary quantum states coupled with the main quantum state, forming a cross-modal quantum coding result; Based on changes in game version, event theme, and operational status, the encoding primitives, coupling weights, and constraint labels in the cross-modal quantum coding results are dynamically updated to generate a dynamic quantum coding library. Based on a dynamic quantum coding library, a QUBO model is constructed with advertising effectiveness and experience consistency as optimization objectives and operational capacity and platform adaptation as dynamic constraints. The QUBO model is solved using quantum annealing or quantum heuristic algorithms to obtain the target combination of advertising creatives; The system decodes and matches the target ad creative mix, and generates corresponding ad delivery plans based on differences in terminal environments.

2. The method according to claim 1, characterized in that, The process of acquiring multimodal data on game content, real-time operational data, user feedback data, and terminal environment data of MMO games, and generating an advertising material feature set based on a unified feature specification, specifically includes: Establish unified feature specifications in advance, and set material identifiers, activity identifiers, region / server identifiers, user group identifiers, and terminal platform identifiers through unified feature specifications; Acquire multimodal data of game content, real-time operational data, user feedback data, and terminal environment data, and associate and bind them based on material identifiers, event identifiers, server identifiers, user group identifiers, and terminal platform identifiers to form material data records; Data preprocessing and feature normalization are performed on the material data records to obtain standardized material features; In accordance with unified feature specifications, standardized material features are organized into advertising material feature sets.

3. The method according to claim 1, characterized in that, The method, based on the feature set of advertising materials, encodes visual content features into a main quantum state and encodes audio features, text semantic features, and user preference features into auxiliary quantum states coupled to the main quantum state, forming a cross-modal quantum coding result, specifically including: Visual content features that characterize the advertiser's visual expression are extracted from the feature set of advertising materials, and the visual content features are discretized and weighted according to the preset visual coding rules to generate the master quantum state; Audio features, text semantic features, and user preference features associated with visual content features are extracted from the feature set of advertising materials, and encoded into auxiliary quantum states according to a unified scale conversion rule; Based on the matching relationship between visual content features and audio features, text semantic features and user preference features, the coupling weight between the master quantum state and the auxiliary quantum state is determined. The master quantum state, auxiliary quantum state, and coupling weights are associated and stored to form a cross-modal quantum encoding result that characterizes the cross-modal matching relationship of advertising materials.

4. The method according to claim 3, characterized in that, The process of dynamically updating the encoding primitives, coupling weights, and constraint labels in the cross-modal quantum encoding results based on changes in game version, event theme, and operational status to generate a dynamic quantum encoding library specifically includes: Obtain game version information, event configuration information, and operation monitoring information, and identify external change events based on the game version information, event configuration information, and operation monitoring information; The affected material range is determined based on external change events, and the master quantum state, auxiliary quantum state and coupling weight corresponding to the material range are retrieved from the cross-modal quantum coding results. Based on the type of change in external events, the encoding primitives corresponding to the master quantum state and auxiliary quantum state are added, replaced, or deactivated, and the coupling weights are adjusted. Based on the conditions of external change events, configure corresponding constraint labels for the cross-modal quantum encoding results; The updated encoding primitives, coupling weights, and constraint labels are indexed and stored to generate a dynamic quantum coding library.

5. The method according to claim 1, characterized in that, The aforementioned QUBO model, based on a dynamic quantum coding library, is constructed with advertising effectiveness and user experience consistency as optimization objectives and operational capacity and platform adaptation as dynamic constraints. Specifically, it includes: Based on the activity identifier, target user group, delivery channel, terminal type, and delivery time window of the current delivery task, the active advertising creative code records are selected from the dynamic quantum coding library, and the selected advertising creative code records are mapped as creative candidate variables; Historical performance data, campaign matching data, and cross-modal consistency data are extracted from advertising creative coding records to generate performance evaluation weights and experience consistency weights. Extract operational constraint parameters and platform adaptation constraint parameters corresponding to server capacity, region / server status, delivery frequency, platform review, ad slot size, and material format from the constraint tags of the dynamic quantum coding library; Target items for prioritizing high-quality materials are set based on performance evaluation weights and experience consistency weights, and penalty items for limiting incompatible materials are set based on operational constraint parameters and platform adaptation constraint parameters. The QUBO model combines candidate variables, target terms, penalty terms, and collaborative or conflicting relationships between materials.

6. The method according to claim 5, characterized in that, The process of extracting historical performance data, campaign matching data, and cross-modal consistency data from advertising creative coding records to generate performance evaluation weights and experience consistency weights specifically includes: Extract historical performance data corresponding to exposure, clicks, conversions, retention, paid subscriptions, recalls, negative feedback, and creative fatigue from the historical delivery logs associated with the advertising creative coding records. Perform anomaly filtering and time decay processing on the historical performance data to obtain a basic advertising performance score. Extract activity matching data related to the current event theme, character profession, gameplay selling points, benefits, and target user group from the advertising creative coding records, and compare the activity matching data with the event configuration information of the current delivery task to adjust the basic score of advertising performance; Consistency data between visual content, audio sentiment, text semantics, and user preferences is extracted from cross-modal quantum coding results in advertising creative coding records, and an experience consistency score is generated based on the consistency data. The performance evaluation weights are generated based on the adjusted basic advertising performance score, and the experience consistency weights are generated through the experience consistency score.

7. The method according to claim 5, characterized in that, The process of combining candidate variables, target terms, penalty terms, and collaborative or conflicting relationships between materials into a QUBO model specifically includes: Generate synergistic relationships based on the complementary relationships between materials in terms of roles, gameplay, benefits, channels, and conversion paths; Conflict relationships are generated based on the incompatibilities between materials in terms of copywriting, visuals, event versions, target audiences, and platform rules; Candidate variables, target terms, penalty terms, synergistic relationships, and conflict relationships are written into a quadratic unconstrained binary optimization structure to form a QUBO model for solving the target ad creative combination.

8. The method according to claim 5, characterized in that, The process of using quantum annealing or quantum heuristic algorithms to solve the QUBO model to obtain the target advertising creative combination specifically includes: Read the candidate variables, target terms, penalty terms, and cooperative or conflict relationships between materials in the QUBO model, and determine the quantum annealing solution mode or quantum heuristic solution mode based on available computing resources; When the quantum annealing solution mode is determined, the QUBO model is converted into an input structure that can be recognized by the quantum annealing device, and the annealing rounds, sampling times and annealing time are set to obtain multiple candidate solutions; When the quantum heuristic solution mode is determined, multiple material combination states are initialized based on the QUBO model, and the material combination states are updated through iterative search to obtain multiple candidate solutions; Multiple candidate solutions are constrained, validated, and ranked based on advertising effectiveness, user experience consistency, operational capacity, and platform compatibility to obtain the ranking result. The candidate solutions that meet the preset conditions in the sorting results are restored to a set of advertising material identifiers, and the target advertising material combination is generated.

9. The method according to claim 1, characterized in that, The process of decoding and matching the target ad creative combination, and generating corresponding ad delivery plans based on differences in terminal environments, specifically includes: Based on the variable mapping relationship in the dynamic quantum coding library, the value results of the candidate variables in the target advertising material combination are decoded to obtain the corresponding advertising material information, which includes advertising material identifier, material version, material type and material resource address; Based on the current campaign theme, target user group, ad placement type, delivery channel, and delivery time window, the decoded ad creative information is verified for content tags, review status, creative validity period, and resource availability to determine the matching creatives. Obtain terminal environment difference information corresponding to terminal operating system, screen ratio, resolution, network status, device performance and platform material specifications, and adapt the display ratio, material format, video bitrate, copy language and landing page path of the matching materials according to the terminal environment difference information; The adapted creative materials are then bound to the delivery channels, ad placement types, display formats, delivery time periods, and exposure frequencies to generate an advertising delivery plan for the corresponding terminal environment.

10. An advertising creative search system based on a quantum heuristic algorithm, characterized in that, The system specifically includes: The data acquisition module is used to acquire multimodal data of game content, real-time operation data, user feedback data and terminal environment data of MMO games, and generate advertising material feature sets based on unified feature specifications; The quantum coding module is used to encode visual content features into the main quantum state based on the feature set of advertising materials, and encode audio features, text semantic features and user preference features into auxiliary quantum states coupled with the main quantum state, forming a cross-modal quantum coding result; The dynamic update module is used to dynamically update the encoding primitives, coupling weights, and constraint labels in the cross-modal quantum coding results based on changes in game version, event theme, and operational status, thereby generating a dynamic quantum coding library. The model building module is used to build a QUBO model based on a dynamic quantum coding library, with advertising effectiveness and experience consistency as optimization goals and operational capacity and platform adaptation as dynamic constraints. The model solving module is used to solve the QUBO model using quantum annealing or quantum heuristic algorithms to obtain the target ad creative combination; The ad delivery module is used to decode and match target ad creative combinations, and generate corresponding ad delivery plans based on differences in terminal environments.

Images