AI model-based reasoning method, communication device, communication equipment and medium

Abstract

The invention discloses a reasoning method based on an AI model in a communication system, a communication device, communication equipment and a medium, and belongs to the field of communication.The method comprises the steps that first equipment obtains first information, and a first model is deployed on the first equipment; according to the first information, the first model is adapted to a target task; wherein the first information comprises at least one of the following information: prompt indication information, knowledge indication information and fine tuning data set indication information.

Description

Technical Field

[0001] This application belongs to the field of communication technologies, and particularly relates to an inference method based on an AI model, a communication device, a communication equipment, and a medium. Background Art

[0002] In a mobile communication system, there are more and more use cases combined with artificial intelligence (AI). For example, at the physical layer, there are AI-based CSI (channel state information) prediction and feedback compression, AI-based beam management, AI-based positioning, etc. In order to make different communication scenarios universal, it is proposed to apply large models to communication networks to solve problems in communication systems. However, large models usually have better performance in solving general problems, but there are many different problems in communication systems, such as beamforming, resource allocation, channel prediction, etc. Therefore, how to ensure the performance of large models in solving specific downstream problems is an urgent problem to be solved. Summary of the Invention

[0003] Embodiments of this application provide an inference method based on an AI model, a communication device, a communication equipment, and a medium, which can improve the model performance when inferring downstream problems based on a large model.

[0004] In a first aspect, an inference method based on an AI model is provided. The method includes:

[0005] A first device obtains first information, where a first model is deployed on the first device; and adapts the first model to a target task according to the first information, where the first information includes at least one of the following information: prompt indication information, knowledge indication information, and fine-tuning dataset indication information.

[0006] In a second aspect, an inference method based on an AI model is provided. The method includes:

[0007] A second device sends first information to the first device, where the first information is used for the first device to adapt the first model to a target task;

[0008] where the first information includes at least one of the following information: prompt indication information, knowledge indication information, and fine-tuning dataset indication information.

[0009] In a third aspect, a communication device is provided, including:

[0010] A processing unit, configured to obtain first information, where a first model is deployed on the communication device; and

[0011] adapt the first model to a target task according to the first information;

[0012] Among them, the first information includes at least one of the following information: prompt indication information, knowledge indication information, and fine-tuning dataset indication information.

[0013] In a fourth aspect, a communication device is provided, including:

[0014] A communication unit, configured to send first information to a first device, where a first model is deployed on the first device, and the first information is used for the first device to adapt the first model to a target task; among them, the first information includes at least one of the following information: prompt indication information, knowledge indication information, and fine-tuning dataset indication information.

[0015] In a fifth aspect, a communication device is provided, the communication device includes a processor and a memory, the memory stores a program or instruction that can run on the processor, and when the program or instruction is executed by the processor, the steps of the method described in any one of the first aspect to the second aspect are implemented.

[0016] In a sixth aspect, a readable storage medium is provided, where a program or instruction is stored on the readable storage medium, and when the program or instruction is executed by a processor, the steps of the method described in any one of the first aspect to the second aspect are implemented.

[0017] In a seventh aspect, a wireless communication system is provided, including: a first device and a second device, the first device can be used to execute the steps of the method described in the first aspect, and the second device can be used to execute the steps of the method described in the second aspect.

[0018] In an eighth aspect, a chip is provided, the chip includes a processor and a communication interface, the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the method described in any one of the first aspect to the second aspect.

[0019] In a ninth aspect, a computer program / program product is provided, the computer program / program product is stored in a storage medium, and the program / program product is executed by at least one processor to implement the steps of the method described in any one of the first aspect to the second aspect.

[0020] In the embodiments of the present application, the first device can obtain auxiliary information for the inference of the first model from the second device, such as prompt indication information, knowledge indication information, fine-tuning datasets, etc. Thus, the inference node can use the first model for inference based on the prompt indication information, knowledge indication information, fine-tuning datasets, etc., which is beneficial to improving the inference performance of the model. Description of the Drawings

[0021] Figure 1It is a schematic diagram of a communication system provided by an embodiment of the present application.

[0022] Figure 2 It is a schematic diagram of an inference method based on an AI model provided by an embodiment of the present application.

[0023] Figure 3 It is a schematic interaction diagram of an inference method based on an AI model provided by an embodiment of the present application.

[0024] Figure 4 It is a schematic interaction diagram of another inference method based on an AI model provided by an embodiment of the present application.

[0025] Figure 5 It is a schematic interaction diagram of yet another inference method based on an AI model provided by an embodiment of the present application.

[0026] Figure 6 It is a schematic diagram of a communication device provided by an embodiment of the present application.

[0027] Figure 7 It is a schematic diagram of another communication device provided by an embodiment of the present application.

[0028] Figure 8 It is a schematic diagram of a communication device provided by an embodiment of the present application.

[0029] Figure 9 It is a hardware structure diagram of a terminal provided by an embodiment of the present application.

[0030] Figure 10 It is a hardware structure diagram of a network-side device provided by an embodiment of the present application.

[0031] Figure 11 It is a hardware structure diagram of another network-side device provided by an embodiment of the present application. Detailed implementation manners

[0032] Next, the technical solutions in the embodiments of the present application will be clearly described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application belong to the scope protected by the present application.

[0033] The terms "first", "second", etc. in this application are used to distinguish similar objects, rather than to describe a specific order or sequence. It should be understood that such terms can be interchanged under appropriate circumstances, so that the embodiments of this application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first" and "second" are generally of the same type, and do not limit the number of objects. For example, the first object can be one or more. In addition, "or" in this application means at least one of the connected objects. For example, "A or B" covers three scenarios, namely, Scenario 1: including A and not including B; Scenario 2: including B and not including A; Scenario 3: including both A and B. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0034] The term "indication" in this application can be either a direct indication (or an explicit indication) or an indirect indication (or an implicit indication). Among them, a direct indication can be understood as that the sender clearly informs the receiver of specific information, operations to be performed, or request results, etc. in the sent indication; an indirect indication can be understood as that the receiver determines the corresponding information according to the indication sent by the sender, or makes a judgment and determines the operations to be performed or request results, etc. according to the judgment result.

[0035] It is worth pointing out that the technology described in the embodiments of this application is not limited to the Long Term Evolution (LTE) / LTE-Advanced (LTE-A) system, and can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), or other systems. The terms "system" and "network" in the embodiments of this application are often used interchangeably, and the described technology can be used not only in the systems and radio technologies mentioned above, but also in other systems and radio technologies. The following description describes the New Radio (NR) system for example purposes, and uses NR terms in most of the following descriptions, but these technologies can also be applied to systems other than the NR system, such as the 6th generation (6 thGeneration, 6G) communication system.

[0036] Figure 1 The block diagram of a wireless communication system to which embodiments of the present application can be applied is shown. The wireless communication system includes a terminal 11 and a network-side device 12. Among them, the terminal 11 can be a mobile phone, a tablet personal computer, a laptop computer, a notebook computer, a personal digital assistant (PDA), a handheld computer, a netbook, an ultra-mobile personal computer (UMPC), a mobile internet device (MID), an augmented reality (AR), a virtual reality (VR) device, a robot, a wearable device, a flight vehicle, a vehicle user equipment (VUE), a shipborne device, a pedestrian user equipment (PUE), a smart home (home appliances with wireless communication functions, such as refrigerators, TVs, washing machines, or furniture, etc.), a game console, a personal computer (PC), an ATM, or a self-service machine, etc., terminal-side devices. Wearable devices include: smart watches, smart bracelets, smart earphones, smart glasses, smart jewelry (smart bracelets, smart bracelets, smart rings, smart necklaces, smart anklets, smart ankle chains, etc.), smart wristbands, smart clothing, etc. Among them, vehicle user equipment can also be referred to as vehicle terminal, vehicle controller, vehicle module, vehicle component, vehicle chip, or vehicle unit, etc. It should be noted that the specific type of the terminal 11 is not limited in the embodiments of the present application.

[0037] The terminal can also be referred to as user equipment (UE), terminal device, access terminal, user unit, user station, mobile station, mobile platform, remote station, remote terminal, mobile device, user terminal, terminal, wireless communication device, user agent, or user device, etc.

[0038] The network-side device 12 may include an access network device or a core network device. Among them, the access network device may also be referred to as a radio access network (RAN) device, a radio access network function, or a radio access network unit. The access network device may include a base station, a wireless local area network (WLAN) access point (AP), or a wireless fidelity (WiFi) node, etc. Among them, the base station may be referred to as Node B (NB), evolved Node B (eNB), next generation Node B (gNB), new radio Node B (NR Node B), access point, relay base station (RBS), serving base station (SBS), base transceiver station (BTS), radio base station, radio transceiver, basic service set (BSS), extended service set (ESS), home Node B (HNB), home evolved Node B, transmission reception point (TRP), or some other suitable term in the art. As long as the same technical effect is achieved, the base station is not limited to specific technical terms. It should be noted that in the embodiments of this application, only the base station in the NR system is taken as an example for introduction, and the specific type of the base station is not limited.

[0039] The core network device may include, but is not limited to, at least one of the following: core network node, core network function, Mobility Management Entity (MME), Access and Mobility Management Function (AMF), Session Management Function (SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Server Discovery Function (EASDF), Unified Data Management (UDM), Unified Data Repository (UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (L-NEF), Binding Support Function (BSF), Application Function (AF), etc. It should be noted that in the embodiments of this application, only the core network devices in the NR system are taken as examples for introduction, and the specific types of core network devices are not limited.but not limited to at least one of the following: core network node, core network function, Mobility Management Entity (MME), Access and Mobility Management Function (AMF), Session Management Function (SMF), User Plane Function (UPF), Policy Control Function (PCF), Policy and Charging Rules Function (PCRF), Edge Application Server Discovery Function (EASDF), Unified Data Management (UDM), Unified Data Repository (UDR), Home Subscriber Server (HSS), Centralized network configuration (CNC), Network Repository Function (NRF), Network Exposure Function (NEF), Local NEF (L-NEF), Binding Support Function (BSF), Application Function (AF), etc. It should be noted that in the embodiments of this application, only the core network devices in the NR system are taken as examples for introduction, and the specific types of core network devices are not limited.

[0040] To facilitate the understanding of the embodiments of this application, the related technologies of this application are described.

[0041] One: Prompt engineering

[0042] Prompt engineering is a technique in Natural Language Processing (NLP) that can create text fragments, prompts, or templates to guide pre-trained large language models to generate high-quality outputs for specific tasks or applications. It is widely used in fields such as question answering, summarization, translation, sentiment analysis, and text generation. Prompt engineering can utilize the powerful capabilities of pre-trained large language models to achieve various complex natural language processing tasks, reduce the dependence on labeled data and model fine-tuning, lower development costs and time, as well as improve the interpretability and controllability of pre-trained large language models, increasing user trust and satisfaction. Common prompts can include Zero-Shot Prompting, Few-Shot Prompting, Chain of Thought Prompting, etc.

[0043] In prompt engineering, the description of the task is embedded in the input to guide the generative AI solution to generate the required output.

[0044] II. Knowledge Graph

[0045] A knowledge graph is a knowledge base that uses a graph structure or topology to represent and integrate data. It can store the associated descriptions between entities (objects, events, situations, or abstract concepts), and at the same time, it can also encode the semantic relationships between entities. The application prospects of knowledge graphs are very broad. It can not only improve the effectiveness of data services such as information retrieval, search engines, and recommendation systems, but also support intelligent interactions such as natural language question answering, dialogue, and reasoning. It has the characteristics of strong representation ability, high flexibility, strong computability, and strong cross-domain ability.

[0046] Knowledge graphs store a large amount of knowledge in a clear and structured way and can be used to enhance the knowledge awareness of large models. Incorporating knowledge graphs into large models during the reasoning stage can significantly improve the performance of large models in accessing domain-specific knowledge by retrieving knowledge from the knowledge graph.

[0047] In the communication field, network data knowledge graphs are mainly used for knowledge representation, correlation analysis, and in-depth mining to provide effective knowledge rules and knowledge computing support for the intelligence of communication systems.

[0048] The network structure, terminal types, terminal behaviors, data service requirements, and system resources of communication systems have characteristics such as high dynamism, strong timeliness, and mutual coupling. Mobile communication data faces many challenges, such as difficult decentralized data acquisition, a wide variety of complex structures, and difficult complex correlation mining, etc.

[0049] Using a knowledge graph can effectively clarify various relationships between data fields and communication network metrics, and further deeply mine based on the established relationships, such as quantifying the degree of association of relationships, characterizing the characteristic attributes of data fields and metrics, etc.

[0050] In some implementation manners, a knowledge graph can be represented by the correlation between entities. One representation can be:

[0051] (Entity 1, correlation coefficient between Entity 1 and Entity 2, Entity 2).

[0052] For example, Entity 1 is the cell throughput, Entity 2 is the L1-RSRP of the strongest beam, and the correlation coefficient is 0.5.

[0053] In some other implementation manners, a knowledge graph can be represented by a Subject, Predicate, Object (SPO) triple. For example, an SPO triple of (cell id, shaping codebook, codebook indication).

[0054] III. Knowledge Vector Library

[0055] A knowledge vector library is a database used to store, retrieve, and analyze vectors. The reason it is called a database is that it has the following characteristics:

[0056] a) Provide a standard access interface to lower the user's usage threshold;

[0057] b) Provide the ability of efficient data organization, retrieval, and analysis. When general users store and retrieve vectors, they also need to manage structured data, that is, support the management ability of traditional databases for structured data.

[0058] Therefore, a knowledge vector library can be simply understood as a database for storing model input feature vectors.

[0059] For example, when using a picture to search for a picture, or using voice to search for voice, what is stored and compared in the knowledge vector library is not the picture and voice segment, but the "features" extracted by algorithms such as deep learning, such as an array of 256 or 512 floating-point numbers (float), which can be represented by vectors in mathematics.

[0060] In some cases, a knowledge vector library can be a model, and the input is a picture, text. For example, in the communication field, the input of the model can be the measurement results of wireless signals, etc., and the output is a feature vector used to characterize the measurement results.

[0061] The following methods can be adopted locally to use the knowledge graph or knowledge vector library:

[0062] 1) Directly used as the input of the model;

[0063] 2) After processing the model input with the knowledge graph, then input the processed data and the original model input into the model together;

[0064] 3) After processing the model input with the knowledge vector library, then input the processed data and the original model input into the model together, or directly use the processed data to input into the model.

[0065] In mobile communication systems, there are more and more use cases combined with Artificial Intelligence (AI). For example, at the physical layer, there are AI-based Channel State Information (CSI) prediction and feedback compression, AI-based beam management, AI-based positioning, etc. In some scenarios, the introduction of AI-based energy saving, AI-based load balancing, etc. is also considered. In the future, there will be more use cases combining AI in mobile communication systems.

[0066] IV. Fine-tuning

[0067] Large language models are first pre-trained on a large-scale unlabeled text dataset, and then fine-tuned on a set of labeled data for the target task. Through fine-tuning, the large language model can better complete the target task.

[0068] In recent years, large language models have received extensive attention in fields such as chatting and image generation. Therefore, it is considered to apply large models to communication networks as a tool to optimize the performance of communication networks. However, there are two problems with communication large models.

[0069] Problem 1: The size of large models is very large. How to deploy large models to devices with limited storage resources such as base stations or terminals for inference? A typical method is to process the large model through compression methods such as quantization and pruning, and then deploy it to communication devices.

[0070] Problem 2: Large models often solve general problems, while there are many different problems in communication, such as beamforming, resource allocation, and channel prediction. How to match large models to specific downstream problems, and there is currently no mature solution in communication systems.

[0071] The following combines the accompanying drawings and details the inference method based on the artificial intelligence AI model in the communication system provided by the embodiments of the present application through some embodiments and their application scenarios.

[0072] Figure 2 Shows a schematic diagram of the inference method based on the artificial intelligence AI model according to the embodiments of the present application. As Figure 2As shown, the method 200 includes:

[0073] S210, a first device obtains first information, where a first model is deployed on the first device;

[0074] S220, adapt the first model to a target task according to the first information.

[0075] In an embodiment of the present application, the first device can be regarded as an inference node, and the first device needs to use the first model to implement the inference of the target task.

[0076] In some embodiments, the first model is a compressed AI model. For example, the first model is an AI model compressed from a large model. Here, the compression may include, but is not limited to, quantization, pruning (for example, deleting some layers in the model, or deleting some nodes in a certain layer), etc.

[0077] It should be noted that in an embodiment of the present application, the AI model can also be referred to as an AI unit, an ML (machine learning) model, an ML unit, an AI structure, an AI function, an AI feature, a machine learning model, a neural network, a neural network function, a neural network function, etc. Or, the AI model can also refer to a processing unit that can implement specific algorithms, formulas, processing flows, capabilities, etc. related to AI, or the AI model can be a processing method, algorithm, function, module or unit for a specific data set, or the AI model can be a processing method, algorithm, function, module or unit running on AI / ML-related hardware such as a Graphics Processing Unit (GPU), a Neural Processing Unit (NPU), a Tensor Processing Unit (TPU), an Application Specific Integrated Circuit (ASIC), etc. The present application does not make specific limitations on this.

[0078] In some embodiments, the target task can be a specific task in a communication system, or a downstream task of the task adapted by the first model. By way of example and not limitation, the target task can be tasks such as beam management, positioning, load balancing, resource allocation, CSI prediction, and compressed feedback.

[0079] In some embodiments, the first information includes at least one of the following information:

[0080] Hint indication information, knowledge indication information, fine-tuning data set indication information.

[0081] In the embodiments of the present application, the prompt indication information can explicitly or implicitly indicate prompt-related information, the knowledge indication information can explicitly or implicitly indicate knowledge-related information, and the fine-tuning dataset indication information can explicitly or implicitly indicate the fine-tuning dataset. The present application does not limit the specific indication method.

[0082] In some embodiments, the prompt-related information and the knowledge-related information can be used to expand the input of the first model and assist the first model in performing accurate reasoning.

[0083] The first device can make the first model better adapted to the target task through the prompt-related information, the knowledge-related information, or the fine-tuning dataset. For example, the first device can prompt or guide the first model to infer an expected output according to the prompt-related information, thereby improving the inference performance of the model. For another example, when performing inference, the first device can combine the knowledge-related information to accurately complete tasks in a professional field and reduce the hallucination problem in model inference. For still another example, the first device can fine-tune the first model with the fine-tuning dataset so that the inference result of the fine-tuned first model conforms to the expected output, thereby improving the inference performance of the model.

[0084] In some embodiments, as Figure 3 shown, S210 may include:

[0085] The first device receives the first information sent by the second device.

[0086] In the embodiments of the present application, the second device can be regarded as a device that assists the first device in performing inference, or an auxiliary inference node. For example, the second device can provide the first device with auxiliary information for obtaining the inference result, or a fine-tuning dataset for fine-tuning the model, so that the first device can adjust the inference result of the model according to the auxiliary information, or fine-tune the model according to the fine-tuning dataset, so that the fine-tuned model can infer an expected output.

[0087] For example, the first device is a terminal, and the second device is an access network device (such as a base station) or a core network function, such as a core network function with a data storage function, such as a database, a data function, or a Network Repository Function (NRF), or a Unified Data Management (UDM), or a third-party server of the terminal, etc.

[0088] For another example, the first device is an access network device, such as a base station, and the second device is a core network function, such as a core network function with a data storage function (such as a database, a data function, an NRF, or a UDM), an AI-related function (such as an AI control function, an AI model management function), operation administration and maintenance, or an access network control node, etc.

[0089] In some embodiments, if the first device is a terminal and the second device is an access network device, the first information is carried in at least one of the following signaling: layer 1 signaling, layer 2 signaling, layer 3 signaling, AI layer signaling, data plane signaling. Exemplarily, the layer 1 signaling may include, but is not limited to, a physical downlink control channel (PDCCH), for example, carrying the first information as downlink control information (DCI) in the PDCCH for transmission. The layer 2 signaling may include, for example, but is not limited to, a downlink media access control control element (MAC CE), and the layer 3 signaling may include, for example, but is not limited to, radio resource control (RRC) signaling.

[0090] In some other embodiments, the first device is a terminal and the second device is a core network function, and the first information is carried in at least one of the following signaling: non-access stratum (NAS) signaling, AI layer signaling, data plane signaling.

[0091] In some embodiments, when the first device is an access network device and the second device is a core network function, the second device may directly send the first information to the first device, or may also forward it to the first device through other nodes (such as a central control node). Optionally, the central control node may include, for example, but is not limited to: an AMF, an AI control node, a task control node, a coordination control node.

[0092] In the embodiments of the present application, the AI control node may be used for the control of AI-related functions; the task control node may be used for the control of service-related functions or for the control of task-related functions. For example, the task may include, but is not limited to, the coordination and allocation of connection, computing, data, and algorithm resources in a multi-node scenario involved in new network capabilities to jointly complete a specific goal; the coordination control node may be used for the coordinated management between multiple functions (such as communication functions, data functions, computing power functions, algorithm functions, model functions).

[0093] In some embodiments of the present application, such as Figure 3 shown, before S210, the method 200 further includes:

[0094] S203, the first device sends second information to the second device.

[0095] In some embodiments, the second information can be used by the first device to request first information from the second device.

[0096] That is, when the second device receives the first information from the first device, it is considered that the first device has a need to obtain the first information. Therefore, the second device can send the second information to the first device based on the first information.

[0097] In some embodiments, the second information includes at least one of the following:

[0098] Requirement information for inference by the first model;

[0099] Auxiliary information for determining the first information;

[0100] Identification information of at least one event that triggers the first device to send the requirement information or the auxiliary information.

[0101] In some embodiments, the first information is determined based on the second information.

[0102] That is, the auxiliary inference node can determine the first information to be sent to the inference node based on the requirement information and / or auxiliary information provided by the inference node, and then send the first information to the first device. Thus, the first device can use the first information for inference based on the first information, which is beneficial to improving the inference performance of the model.

[0103] For example, the second device can select appropriate prompt indication information, knowledge-related information, or fine-tuning data sets according to the requirement information to meet the inference requirements of the first model.

[0104] For another example, the second device can select appropriate prompt indication information, knowledge-related information, or fine-tuning data sets according to the auxiliary information to assist the first device in using the first model to infer an output that meets the expectations.

[0105] In some embodiments, if the first device is a terminal and the second device is an access network device, the second information is carried in at least one of the following signaling: layer 1 signaling, layer 2 signaling, layer 3 signaling, AI layer signaling, data plane signaling. Exemplarily, the layer 1 signaling may include, but is not limited to, a Physical Uplink Control Channel (PUCCH). For example, the second information is carried as Uplink Control Information (UCI) in the PUCCH for reporting. The layer 2 signaling may include, for example, but is not limited to, an uplink MAC CE, and the layer 3 signaling may include, for example, but is not limited to, RRC signaling.

[0106] In some other embodiments, the first device is a terminal and the second device is a core network function, and the second information is carried in at least one of the following signaling: NAS signaling, AI layer signaling, data plane signaling.

[0107] In some embodiments, when the first device is an access network device and the second device is a core network function, the first device may directly send the second information to the second device, or may also forward it to the second device through other nodes (such as a central control node). Optionally, the central control node may include, for example, but is not limited to: an AMF, an AI control node, a task control node, a collaborative control node.

[0108] In some embodiments, the requirement information for the first model to perform inference includes, but is not limited to, at least one of the following:

[0109] An indication of the target task to be adapted by the AI model;

[0110] The quality of experience (QoE) requirement for the AI model's service;

[0111] The service processing latency of the AI model;

[0112] The service processing accuracy of the AI model;

[0113] The service computing volume of the AI model;

[0114] The data processing scale of the AI model.

[0115] For example, the second device may select hint indication information, knowledge-related information, or a fine-tuning data set related to the target task according to the target task indication.

[0116] Again, for example, the second device may select hint indication information, knowledge-related information, or a fine-tuning data set that meets the QoE requirement, or processing latency, processing accuracy, or processing scale, according to the QoE requirement, or processing latency, processing accuracy, or processing scale.

[0117] In some embodiments, the auxiliary information for determining the first information includes, but is not limited to, at least one of the following:

[0118] The identification information of the AI model, the function indication of the AI model, the characteristic indication of the AI model, and the configuration indication of the first device.

[0119] For example, the second device may select the hint indication information, knowledge-related information, or fine-tuning data set related to the function (or characteristic) indicated by the function indication (or characteristic indication).

[0120] For another example, the second device may select the knowledge-related information in the field related to the configuration indication according to the configuration indication of the first device, or the fine-tuning data set applicable to the configuration indication.

[0121] In some embodiments, the first device is a terminal, and the second information may be sent under specific circumstances, for example:

[0122] When at least one of the following events occurs, the first device sends the second information to the second device:

[0123] Event 1: Receiving a first indication for indicating activation or switching of the model;

[0124] Event 2: The first device determines to activate or switch the model;

[0125] Event 3: The first device determines that the currently used model does not meet the requirements;

[0126] Event 4: Receiving a second indication for indicating that the currently used model does not meet the requirements.

[0127] In some embodiments, the event that triggers the first device to send the second information to the second device may be predefined or configured by the network-side device. When the event that triggers the second information is satisfied, the first device may send the identification information of the event to the network-side device, for example, sending it to the network-side device carried in the second information, so that the network-side device can learn about the events occurring on the first device. Optionally, the first device may also send the characteristic indication and / or function indication of the AI model to the second device. That is, the second information may include the identification information of the event and the characteristic indication and / or function indication of the AI model. The characteristic indication may be used to indicate the characteristics that the first device expects the AI model to achieve, and the function indication may be used to indicate the functions that the first device expects the AI model to achieve.

[0128] In some embodiments, when at least one of the above events occurs, the first device may also not send the auxiliary information and / or requirement information to the second device, but only send the identification information of the event that occurs on the first device to the second device. Optionally, the first device may further send the characteristic indication and / or function indication of the AI model to the second device to implicitly indicate to the second device the requirement for obtaining the first information. Further, the second device may determine the corresponding first information according to the identification information of the event and send it to the first terminal.

[0129] Therefore, in the embodiments of the present application, when at least one of the above events occurs, the first device sends the second information to the second device, which is beneficial for the second device to timely indicate the first information to the first device, so that the first device can use the first information to expand the input of the first model to assist in the output of the first model, or adjust the first model, so that the output of the first model can match the target task.

[0130] In some embodiments, the first indication may be sent by a network-side device. For example, the network-side device may control the activation or switching of the model on the first device. For example, when the model currently used on the first device does not meet the requirements, the network-side device may send the first indication to the first device, so that the first device can activate the first model or switch to the first model according to the first indication.

[0131] In some embodiments, when the performance of the model currently used by the first device does not meet the requirements, the first device may independently determine to activate the first model or switch to the first model.

[0132] In some embodiments, the second indication may be sent by a network-side device. For example, when the model currently used on the first device does not meet the requirements, the network-side device may send the second indication to the first device to indicate that the model currently used by the first device does not meet the requirements. Further, the first device may independently determine whether to activate the first model or switch to the first model according to the second indication.

[0133] In some embodiments, the model not meeting the requirements may mean that the model does not meet the requirement information, for example, it may include but is not limited to at least one of the following: the inference of the model does not meet the quality of experience (QoE) requirement, the service processing delay of the model does not meet the delay requirement, and the service processing accuracy of the model does not meet the accuracy requirement.

[0134] In some embodiments of the present application, as Figure 3 shown, the method 200 further includes:

[0135] S202, the first device receives the first model sent by the third device.

[0136] Since the first device needs to implement a relatively large number of communication tasks and the storage space of the first device is limited, if all AI models are stored on the first device, it will consume a large amount of the storage space of the first device. Therefore, the first model may not be stored on the first device. When it is necessary to use the first model to implement corresponding tasks, the first model is obtained from the third device, and then the first model is further used to implement the corresponding tasks.

[0137] In the embodiments of the present application, the third device may be considered as a storage device for models, or, model storage node, or, AI model library, or, AI model management function.

[0138] Optionally, the second device and the third device may be the same device, or, may also be different devices. That is, the auxiliary inference node and the model storage node may be the same device, or, may also be different devices.

[0139] In some embodiments, the first device is a terminal, and the third device is an access network device (such as a base station) or a core network function, such as a core network function with data storage function, such as a database, data function or Network Repository Function (NRF) or UDM, or an AI model library, or an AI model management function or a third-party server of the terminal, etc.

[0140] In other embodiments, the first device is an access network device, and the third device is a core network function, for example, a core network function with data storage function (such as a database, data function or NRF or UDM), AI-related functions (such as AI control function, AI model management function, AI model library), Operation Administration and Maintenance or a third-party server of the access network device, etc.

[0141] In some embodiments, if the first device is a terminal and the third device is an access network device, the first model is carried in at least one of the following signaling: layer 1 signaling, layer 2 signaling, layer 3 signaling, data plane signaling, AI layer signaling. Exemplarily, the layer 1 signaling may include, but is not limited to, PDCCH. The layer 2 signaling may include, for example, but is not limited to, downlink MAC CE, and the layer 3 signaling may include, for example, but is not limited to, RRC signaling.

[0142] In other embodiments, the first device is a terminal. If the third device is a core network function, the first model is carried in at least one of the following signaling: NAS signaling, data plane signaling, AI layer signaling.

[0143] In some embodiments of the present application, as Figure 3As shown, before S202, the method 200 further includes:

[0144] S201, the first device sends third information to the third device, and the third information may include model-related information.

[0145] In some embodiments, if the first device is a terminal and the third device is an access network device, the third information is carried in at least one of the following signaling: layer 1 signaling, layer 2 signaling, layer 3 signaling, data plane signaling, AI layer signaling. Exemplarily, the layer 1 signaling may include but is not limited to PUCCH. For example, the second information is carried as UCI in PUCCH for reporting. The layer 2 signaling may include but is not limited to uplink MAC CE, and the layer 3 signaling may include but is not limited to RRC signaling.

[0146] In some embodiments, the first device is a terminal. If the third device is a core network function, the third information is carried in at least one of the following signaling: NAS signaling, data plane signaling, AI layer signaling.

[0147] In some embodiments, the third information includes but is not limited to at least one of the following:

[0148] Identification information of the AI model, function indication of the AI model, characteristic indication of the AI model, configuration indication of the first device.

[0149] In some embodiments, the first model is selected based on the third information. For example, the first model is a model corresponding to the identification information, a model that meets the function indication, or a model that meets the specific indication, or a model that meets the configuration indication. Therefore, the third device selects a suitable model according to the third information and sends it to the first device, which is beneficial to meeting the inference requirements of the first device.

[0150] In some embodiments, the identification information of the AI model may be, for example, the model identification (modelID) of the AI model, which can be used for AI structure identification, AI algorithm identification, or the identification of a specific dataset associated with the AI model, or the identification of a specific scenario, environment, channel characteristic, or device related to AI / ML, or the identification of a function, characteristic, ability, or module related to AI / ML. The present invention does not make specific limitations thereto.

[0151] In some embodiments, the characteristic indication of the AI model may explicitly or implicitly indicate the features supported by the AI model, such as supporting CSI prediction and compression feedback, beam prediction, positioning, load balancing, resource allocation, etc.

[0152] In the embodiments of the present application, it can be considered that the characteristics of the AI model under specific configurations are the functions of the AI model.

[0153] For example, the feature of the AI model is beam prediction, and the function of the AI model is time-domain beam prediction under the 32 transmit beams configured by the base station.

[0154] In some embodiments, the function indication of the AI model may explicitly or implicitly indicate the functions supported by the AI model, such as supporting CSI prediction and compressed feedback, beam prediction, positioning, load balancing, resource allocation, etc. under specific configurations.

[0155] In some embodiments, the configuration indication of the first device includes at least one of the following:

[0156] The number of transmit antennas, the number of transmit beams, the number of antenna ports, the transmit power, the antenna gain, the 3dB bandwidth of the beam, the spacing, the frequency, the system bandwidth.

[0157] Hereinafter, in combination with specific embodiments, the specific content of each piece of information in the first information and the second information will be described.

[0158] In some embodiments, the prompt indication information is related to the target task.

[0159] In some implementation manners, the prompt indication information explicitly indicates the prompt-related information related to the target task.

[0160] Exemplarily, the prompt indication information includes at least one of the following:

[0161] The chain of thought related to the target task;

[0162] The prompt text related to the target task;

[0163] The prompt file related to the target task;

[0164] The prompt code related to the target task.

[0165] In some embodiments, the chain of thought related to the target task may include one or more of a plurality of candidate chains of thought, and the plurality of candidate chains of thought may be predefined or indicated by the network-side device. For example, when the first model needs to be matched to the target task, the second device may select one or more chains of thought from the plurality of candidate chains of thought and indicate them to the first device.

[0166] In some embodiments, the prompt text related to the target task may include one or more of a plurality of candidate prompt texts, and the plurality of candidate prompt texts may be predefined or indicated by the network-side device. For example, the second device may select one or more prompt texts from the plurality of candidate prompt texts and indicate them to the first device.

[0167] In some embodiments, the prompt file associated with the target task may include one or more of a plurality of candidate prompt files, and the plurality of candidate prompt files may be predefined or indicated by a network-side device. For example, the second device may select one or more prompt files from the plurality of candidate prompt files and indicate them to the first device.

[0168] In some embodiments, the prompt code associated with the target task may include one or more of a plurality of candidate prompt codes, and the plurality of candidate prompt codes may be predefined or indicated by a network-side device. For example, the second device may select one or more prompt codes from the plurality of candidate prompt codes and indicate them to the first device.

[0169] In some other implementations, the prompt indication information implicitly indicates prompt related information related to the target task.

[0170] Exemplarily, the prompt indication information includes but is not limited to at least one of the following:

[0171] Thinking chain identification related to the target task;

[0172] The prompt text mark related to the target task;

[0173] The prompt file identifier related to the target task;

[0174] Prompt code identification related to the target task;

[0175] The address of the prompt file related to the target task.

[0176] Optionally, when the first information is transmitted via NAS signaling, the prompt indication information may include a prompt file address related to the target task.

[0177] In some embodiments, multiple thought chain identifiers may be predefined or preconfigured (for example, pre-indicated by a network-side device), each thought chain identifier corresponding to a thought chain. For example, the second device may select one or more thought chain identifiers from the multiple thought chain identifiers and indicate them to the first device. Further, the first device may determine the corresponding thought chain according to the indicated thought chain identifier, and further perform reasoning based on the thought chain.

[0178] In some embodiments, multiple prompt text identifiers may be predefined or preconfigured (for example, pre-indicated by a network-side device), each prompt text identifier corresponding to a prompt text. For example, the second device may select one or more prompt text identifiers from the multiple prompt text identifiers and indicate them to the first device. Further, the first device may determine the corresponding prompt text according to the indicated prompt text identifier, and further perform reasoning based on the prompt text.

[0179] ​In some embodiments, multiple hint file identifiers may be predefined or preconfigured (e.g., pre-indicated by a network-side device). Each hint file identifier corresponds to a hint file. For example, the second device may select one or more hint file identifiers from the multiple hint file identifiers and indicate them to the first device. Further, the first device may determine the corresponding hint file according to the indicated hint file identifier and further perform reasoning based on the hint file.

[0180] In some embodiments, multiple hint code identifiers may be predefined or preconfigured (e.g., pre-indicated by a network-side device). Each hint code identifier corresponds to a segment of hint code. For example, the second device may select one or more hint code identifiers from the multiple hint code identifiers and indicate them to the first device. Further, the first device may determine the corresponding hint code according to the indicated hint code identifier and further perform reasoning based on the hint code.

[0181] In some embodiments, multiple hint file addresses may be predefined or preconfigured (e.g., pre-indicated by a network-side device). Each hint file address corresponds to a hint file. For example, the second device may select one or more hint file addresses from the multiple hint file addresses and indicate them to the first device. Further, the first device may obtain the corresponding hint file according to the indicated hint file address and further perform reasoning based on the hint file.

[0182] In some embodiments, the target task is beam prediction, and the hint indication information may be:

[0183] When the input of the model is the beam quality of 8 fixed beam patterns in the historical 4 time units, the output of the model is the strongest beam identifier in the future 1 time unit. For example, the period of the time unit is 40 ms.

[0184] In the embodiments of the present application, the knowledge indication information may explicitly or implicitly indicate knowledge-related information (such as knowledge graph-related information, knowledge vector library-related information). For example, the knowledge indication information includes at least one of knowledge graph indication information and knowledge vector library indication information.

[0185] In some embodiments, the knowledge indication information is related to the target task. For example, the knowledge indication information is used to indicate the knowledge in the field to which the target task belongs or the scenario corresponding to the target task, such as a knowledge graph, a knowledge vector library, etc.

[0186] In some implementation manners, the knowledge graph indication information may explicitly indicate knowledge graph-related information.

[0187] Exemplarily, the knowledge graph indication information includes but is not limited to at least one of the following:

[0188] Relationships between entities in the knowledge graph;

[0189] Attribute information of entities in the knowledge graph;

[0190] Data structure of the knowledge graph;

[0191] Correlation between entities in the knowledge graph;

[0192] Subject-Predicate-Object (SPO) triples of statements in the knowledge graph;

[0193] Knowledge graph file.

[0194] In some embodiments, the relationships between entities in the knowledge graph can be represented in the following format:

[0195] (Entity 1, Relationship, Entity 2).

[0196] Optionally, the relationships between entities can include but are not limited to hyponymy relationships, inheritance relationships, etc.

[0197] In some embodiments, the attribute information of entities in the knowledge graph can be represented in the following format:

[0198] (Entity, Attribute Name, Attribute Value).

[0199] In some embodiments, the data structure of the knowledge graph can be used to describe the nodes in the knowledge graph and the relationships between the nodes. For example, the relationships between the nodes can be represented by triples, such as a relationship can be (Node 1, Edge, Node 2).

[0200] In some embodiments, the correlation between entities in the knowledge graph can be represented in the following format:

[0201] (Entity 1, Correlation Coefficient between Entity 1 and Entity 2, Entity 2).

[0202] For example, Entity 1 is the cell throughput, Entity 2 is the L1-RSRP of the strongest beam, and the correlation coefficient is 0.8.

[0203] In some embodiments, the SPO triples in the knowledge graph are used to describe the subject, predicate, and object of a statement.

[0204] In some other implementation manners, the knowledge graph indication information can implicitly indicate knowledge graph related information.

[0205] Exemplarily, the knowledge graph indication information includes but is not limited to at least one of the following:

[0206] Knowledge graph identifier;

[0207] Knowledge graph file address.

[0208] Optionally, when the first device is a terminal and the second device is a core network function, the knowledge graph file address can be transmitted through NAS signaling.

[0209] In some embodiments, multiple knowledge graph identifiers can be predefined or preconfigured (e.g., pre-indicated by a network-side device), and each knowledge graph identifier corresponds to a knowledge graph. For example, the second device can select one or more knowledge graph identifiers from the multiple knowledge graph identifiers and indicate them to the first device. Further, the first device can determine the corresponding knowledge graph based on the indicated knowledge graph identifier and further perform reasoning based on the knowledge graph.

[0210] In some embodiments, multiple knowledge graph file addresses can be predefined or preconfigured (e.g., pre-indicated by a network-side device), and each knowledge graph file address corresponds to a knowledge graph file. For example, the second device can select one or more knowledge graph file addresses from the multiple knowledge graph file addresses and indicate them to the first device. Further, the first device can obtain the corresponding knowledge graph file based on the indicated knowledge graph file address and further perform reasoning based on the knowledge graph file.

[0211] In some embodiments, the target task is beam prediction, and the usage mode of the knowledge graph can be:

[0212] According to the cell identifier and the target task, the cell transmission beam configuration, such as the pointing of the transmission beam or the shaping codebook, can be retrieved using the knowledge graph.

[0213] In some implementation manners, the knowledge vector library indication information can implicitly indicate the knowledge vector library related information.

[0214] Exemplarily, the knowledge vector library indication information includes but is not limited to at least one of the following:

[0215] Knowledge vector library identifier;

[0216] Knowledge vector library file address.

[0217] Optionally, when the first information is transmitted through NAS signaling, the knowledge indication information can include the knowledge vector library file address related to the target task.

[0218] In some embodiments, multiple knowledge vector library identifiers may be predefined or preconfigured (e.g., pre-indicated by a network-side device), and each knowledge vector library identifier corresponds to a knowledge vector library. For example, the second device may select one or more knowledge vector library identifiers from the multiple knowledge vector library identifiers and indicate them to the first device. Further, the first device may determine the corresponding knowledge vector library according to the indicated knowledge vector library identifier, and further perform reasoning based on the knowledge vector library.

[0219] In some embodiments, multiple knowledge vector library file addresses may be predefined or preconfigured (e.g., pre-indicated by a network-side device), and each knowledge vector library file address corresponds to a knowledge vector library file. For example, the second device may select one or more knowledge vector library file addresses from the multiple knowledge vector library file addresses and indicate them to the first device. Further, the first device may obtain the corresponding knowledge vector library file according to the indicated knowledge vector library file address, and further perform reasoning based on the knowledge vector library file.

[0220] In some embodiments, the target task is beam prediction, and the usage method of the knowledge vector library may be as follows:

[0221] According to the cell identifier and the target task, use the knowledge vector library to obtain the cell transmission beam configuration, such as the pointing of the transmission beam or the eigenvector related to the shaping codebook.

[0222] In some embodiments, the fine-tuning dataset indication information is used to indicate the input information and label information of the first model, and the input information of the first model is related to the target task. Therefore, the first device trains the first model based on the input information and label information of the first model to achieve fine-tuning of the first model, so that the fine-tuned first model is adapted to the target task.

[0223] In some embodiments, the fine-tuning dataset indication information may explicitly indicate the fine-tuning dataset. For example, the fine-tuning dataset indication information includes the input information and label information of the first model. Alternatively, it may also implicitly indicate the fine-tuning dataset. For example, the fine-tuning dataset indication information includes the identifier of the fine-tuning dataset, and this identifier corresponds to a fine-tuning dataset.

[0224] In some embodiments, the fine-tuning dataset may be determined according to the datasets reported by at least one terminal.

[0225] Optionally, considering data privacy, the sensitive data in the fine-tuning dataset may be privatized. For example, the data in a fine-tuning data may be data obtained by using the same privatization processing method.

[0226] In some embodiments, the input information of the first model includes but is not limited to at least one of the following:

[0227] Measurement results of wireless signals, sensing results, and wireless statistical characteristics.

[0228] In some embodiments, the measurement results of wireless signals may be measurement results obtained by measuring wireless signals, and the wireless signals may include, for example, but are not limited to, uplink signals (such as uplink reference signals), downlink signals (such as downlink reference signals), and sidelink signals (such as sidelink reference signals).

[0229] Exemplarily, the measurement results of the wireless signals may include, for example, but are not limited to:

[0230] Rank Indication (RI);

[0231] Precoding Matrix Indicator (PMI);

[0232] Channel Quality Indicator (CQI)

[0233] Reference Signal Receiving Quality (RSRQ);

[0234] Reference Signal Receiving Power (RSRP);

[0235] Signal to Interference plus Noise Ratio (SINR);

[0236] Signal to Noise Ratio (SNR)

[0237] Received Signal Strength Indication (RSSI).

[0238] In some embodiments, the sensing results are obtained by sensing wireless signals.

[0239] In some embodiments, the sensing results obtained from wireless signals include at least one of the following:

[0240] The first-level measurement quantity (representing the received signal / original channel information). For example, it includes but is not limited to: the complex result of the received signal / channel response, amplitude / phase, I-channel / Q-channel and their operation results (the operations include addition, subtraction, multiplication, division, matrix addition, subtraction, multiplication, matrix transpose, trigonometric relation operations, square root operation, power operation, etc., as well as the threshold detection results and maximum / minimum value extraction results of the above operation results; the operations also include Fast Fourier Transform (FFT) / Inverse Fast Fourier Transform (IFFT), Discrete Fourier Transform (DFT) / Inverse Discrete Fourier Transform (IDFT), 2D-FFT, 3D-FFT, matched filtering, autocorrelation operation, wavelet transform, digital filtering, etc., as well as the threshold detection results and maximum / minimum value extraction results of the above operation results);

[0241] The second-level measurement quantity (basic measurement quantity), including: time delay, Doppler, angle, intensity, and their multi-dimensional combined representation;

[0242] The third-level measurement quantity (basic attribute / status), including but not limited to: distance, speed, orientation, spatial position, acceleration;

[0243] The fourth-level measurement quantity (advanced attribute / status), including but not limited to: whether the target exists, trajectory, action, expression, vital signs, quantity, imaging result, weather, air quality, shape, material, composition.

[0244] In some other embodiments, the perception result can also be obtained through a third-party device.

[0245] Exemplarily, the third-party device includes at least one of the following:

[0246] Camera, radar, lidar, Global Positioning System (GPS), inertial measurement unit.

[0247] In some embodiments, the perception results related to lidar include at least one of the following:

[0248] Lidar point cloud data, and each point in the lidar point cloud data includes: X / Y / Z position information, additional information;

[0249] The angle and distance of the target obtained from the lidar point cloud data;

[0250] The vision-related measurement quantities of the target identified from the lidar point cloud data, such as: people, vehicles, etc.;

[0251] The number of targets identified from the lidar point cloud data.

[0252] In some embodiments, the additional information in the lidar point cloud data includes at least one of the following:

[0253] Intensity: The echo intensity of the laser pulse that generates the lidar point;

[0254] Number of echoes: The number of echoes is the total number of echoes of a given pulse;

[0255] Point classification: Each post-processed lidar point can have a classification that defines the type of object that reflects the lidar pulse. The lidar points can be divided into many categories, such as: ground, bare ground surface, top of the tree canopy, and water area, etc.;

[0256] Red, green, blue (RGB): The RGB band can be used as an attribute of the lidar data, and this attribute usually comes from the image collected during the lidar measurement;

[0257] Global Positioning System (GPS) time: The GPS timestamp of the laser point emitted from the aircraft;

[0258] Scanning angle:

[0259] Scanning direction: The traveling direction of the laser scanning mirror, where the value 1 represents the positive scanning direction and the value 0 represents the negative scanning direction.

[0260] In some embodiments, the vision-related measurement quantities of the targets identified from the lidar point cloud data include at least one of the following:

[0261] Visual image;

[0262] Luminance of the image pixel;

[0263] RGB value of the image pixel;

[0264] Visual features of the targets identified from the image, such as: people, vehicles, etc.

[0265] Angle and distance of the targets identified from the image (especially for binocular vision);

[0266] The number of targets identified from the image.

[0267] In some embodiments, the radar-related perception results include at least one of the following:

[0268] Radar point cloud, and each point in the point cloud includes at least one of: distance / speed / azimuth / pitch angle, or at least one of X / Y / Z / speed;

[0269] Distance, speed, and angle of the identified target;

[0270] Radar imaging;

[0271] Number of targets.

[0272] In some embodiments, the perception results related to the inertial measurement unit include at least one of the following:

[0273] Acceleration: at least one of the three directions of X / Y / Z;

[0274] Speed: at least one of the three directions of X / Y / Z;

[0275] Angular velocity: around at least one of the three axes of X / Y / Z.

[0276] In some embodiments, the perception results related to other third-party devices may include at least one of the following:

[0277] Whether a target exists, its trajectory, actions, expressions, vital signs, quantity, imaging results, weather, air quality, shape, material, composition, etc.

[0278] In some embodiments, the wireless statistical characteristics include at least one of the following:

[0279] Blockage probability, user status, user behavior, handover failure rate.

[0280] Among them, the blockage probability may refer to the probability that the serving beam is blocked.

[0281] In some embodiments, the user status includes at least one of the following:

[0282] Registration Management (RM) status, Connection Management (CM) status, Radio Resource Control (RRC) connection management status.

[0283] Exemplarily, the RM status may include the RM DEREGISTERED status and the RM REGISTERED status.

[0284] Exemplarily, the CM status may include the CM_IDLE status and the CM_CONNECTED status.

[0285] Exemplarily, the RRC connection management state may include the RRC-connected state, the RRC-idle state, and the RRC-inactive state.

[0286] In some embodiments, the user behavior includes at least one of the following:

[0287] Behavior in the RRC-idle state, behavior in the RRC-inactive state, and behavior in the RRC-connected state.

[0288] Exemplarily, the behavior in the RRC-idle state includes at least one of the following:

[0289] Public Land Mobile Network (PLMN) selection, neighbor cell measurement, cell selection, cell reselection, Tracking Area (TA) update, paging monitoring, and obtaining system information.

[0290] Exemplarily, the behavior in the RRC-inactive state includes at least one of the following:

[0291] Neighbor cell measurement, cell reselection, cell selection, RAN notification area (RNA) update, RAN paging monitoring, and obtaining system information.

[0292] Exemplarily, the behavior in the RRC-connected state includes at least one of the following:

[0293] Serving cell channel quality measurement and reporting, neighbor cell measurement and measurement report reporting, PDCCH monitoring, monitoring the control channel related to the shared data channel (detecting whether there is a relevant scheduling), and obtaining system information.

[0294] In some embodiments, taking beam prediction as an example of the target task, the model inputs in the fine-tuning dataset may include the following information:

[0295] Wireless measurement quantity: L1-RSRP, and its transmission beam identifier;

[0296] Perception information: 2D / 3D map of the wireless environment collected by the camera;

[0297] Wireless statistical characteristics: occlusion probability.

[0298] In some embodiments of the present application, when the first device is a terminal, the method 200 further includes:

[0299] The terminal sends at least one of the fourth information and the first capability information to the network-side device, where the fourth information is used to indicate auxiliary information required for the terminal to perform inference based on the first model, and the first capability information is used to indicate the terminal's ability to perform inference based on the first model.

[0300] In some embodiments, the fourth information or the first capability information can be used by the network-side device to determine the first information to be indicated to the terminal.

[0301] Exemplarily, the fourth information includes but is not limited to at least one of the following:

[0302] Indication information for a prompt adapted to the target task;

[0303] Indication information for knowledge adapted to the target task;

[0304] Indication information for a fine-tuning data set adapted to the target task.

[0305] Optionally, the indication information for a prompt adapted to the target task can be used to indicate the prompts expected to be used by the terminal, and this indication information can indicate one or more prompts. For example, the network-side device can select one or more prompts from the prompts expected to be used by the terminal and indicate them to the terminal through the first information.

[0306] Optionally, the indication information for a prompt adapted to the target task can explicitly or implicitly indicate the prompt adapted to the target task.

[0307] Exemplarily, the indication information for a prompt adapted to the target task includes but is not limited to at least one of the following:

[0308] The thought chain adapted to the target task;

[0309] The prompt text adapted to the target task;

[0310] The prompt file adapted to the target task;

[0311] The prompt code adapted to the target task;

[0312] The thought chain identifier adapted to the target task;

[0313] The prompt text identifier adapted to the target task;

[0314] The prompt file identifier adapted to the target task;

[0315] The prompt code identifier adapted to the target task;

[0316] The address of the prompt file adapted to the target task.

[0317] Optionally, the indication information for the knowledge adapted to the target task can be used to indicate the knowledge expected to be used by the terminal, and the indication information can indicate one or more pieces of knowledge. For example, the network-side device can select one or more pieces of knowledge from the knowledge expected to be used by the terminal and indicate them to the terminal through the first information.

[0318] Optionally, the indication information for the knowledge adapted to the target task can explicitly or implicitly indicate the knowledge adapted to the target task.

[0319] In some embodiments, the indication information for the knowledge adapted to the target task can include the indication information for the knowledge graph adapted to the target task and / or the indication information for the knowledge vector library. For example, the indication information for the knowledge graph is one or more knowledge graph identifiers, or a knowledge graph group identifier, a knowledge graph file address, etc., or explicitly indicates one or more knowledge graphs. For example, the indication information for the knowledge vector library is used to indicate one or more knowledge vector library identifiers, or explicitly indicates one or more knowledge vector libraries.

[0320] Exemplarily, the indication information for the knowledge graph includes at least one of the following:

[0321] The relationships between entities in the knowledge graph;

[0322] The attribute information of entities in the knowledge graph;

[0323] The data structure of the knowledge graph;

[0324] The correlation between entities in the knowledge graph;

[0325] The SPO triples of statements in the knowledge graph;

[0326] The knowledge graph file;

[0327] The knowledge graph identifier;

[0328] The knowledge graph file address.

[0329] Optionally, the indication information for the fine-tuning data set adapted to the target task can be used to indicate the fine-tuning data set expected to be used by the terminal, and the indication information can indicate one or more fine-tuning data sets. For example, the network-side device can select one or more fine-tuning data sets from the knowledge expected to be used by the terminal and indicate them to the terminal through the first information.

[0330] Optionally, the fine-tuning data set indicated by the indication information can include at least one of a plurality of predefined or pre-configured fine-tuning data sets. Among them, each fine-tuning data set can include a set of input information and label information.

[0331] Optionally, the indication information for the fine-tuning data set adapted to the target task may explicitly or implicitly indicate one or more fine-tuning data sets adapted to the target task. For example, the indication information may indicate the identification information of the fine-tuning data set, or explicitly indicate one or more fine-tuning data sets.

[0332] Exemplarily, the first capability information includes, but is not limited to, at least one of the following:

[0333] The identification information of the first model supported by the terminal;

[0334] Whether the terminal supports adapting the first model to the target task based on the prompt information;

[0335] Whether the terminal supports adapting the first model to the target task based on the fine-tuning data set;

[0336] Whether the terminal supports adapting the first model to the target task based on the knowledge information;

[0337] The indication information of the prompt supported by the terminal;

[0338] The indication information of the knowledge supported by the terminal;

[0339] The indication information of the fine-tuning data set supported by the terminal.

[0340] Among them, the specific implementation of the indication information of the prompt supported by the terminal refers to the description of the indication information of the prompt adapted to the target task, the specific implementation of the indication information of the knowledge supported by the terminal refers to the description of the indication information of the knowledge adapted to the target task, and the specific implementation of the indication information of the fine-tuning data set supported by the terminal refers to the description of the indication information of the fine-tuning data set adapted to the target task. For the sake of brevity, it will not be elaborated here.

[0341] In some embodiments, the first model may be considered as a model for adapting to the target task, and the identification information of the first model supported by the terminal may be the identification information of the model supported by the terminal that can adapt to the target task.

[0342] In some embodiments, the first capability information may be used by the network-side device to determine whether to indicate the first information to the terminal and / or the content of the first information sent to the terminal.

[0343] For example, when the terminal supports adapting the first model to the target task based on hint information, hint indication information may be sent to the terminal. Or, when the terminal supports adapting the first model to the target task based on knowledge information, hint knowledge indication information may be sent to the terminal. Or, when the terminal supports adapting the first model to the target task based on a fine-tuning data set, hint fine-tuning data set indication information may be sent to the terminal.

[0344] For another example, when the terminal supports multiple hint information, the network-side device may select target hint information from the multiple hint information and send it to the terminal through the first information. Or, when the terminal supports multiple knowledge information, the network-side device may select target knowledge information from the multiple knowledge information and send it to the terminal through the first information. Or, when the terminal supports multiple fine-tuning data sets, the network-side device may select target fine-tuning data sets from the multiple fine-tuning data sets and send them to the terminal through the first information.

[0345] In some embodiments, the terminal sends at least one of the fourth information and the first capability information to the network-side device during the model registration (or model identification) process or the capability reporting process.

[0346] It should be understood that the fourth information and the first capability information may be reported to the network-side device through one signaling, or may also be reported to the network-side device through different signaling. This application does not make a limitation on this.

[0347] Hereinafter, Figures 4 to 5 illustrated by specific embodiments, the inference method based on the AI model provided by this application is described.

[0348] In Figure 4 example, the first device is a terminal, the second device is an access network device or a core network function, the third device is an access network device or a core network function, and the first model is deployed on the first device.

[0349] As Figure 4 shown, the inference method may include the following steps:

[0350] S403. The terminal sends the second information to the access network device or the core network function.

[0351] Among them, the second information includes requirement information for the first model to perform inference and / or auxiliary information for determining the first information.

[0352] S404. The access network device or the core network function sends the first information to the terminal. Among them, the first information is determined according to the second information.

[0353] Among them, the first information includes at least one of prompt indication information, knowledge indication information, and fine-tuning dataset indication information.

[0354] In some embodiments, before S403, the method further includes:

[0355] S401, the terminal may send third information to the access network device or the core network function;

[0356] S402, the terminal receives the first model sent by the access network device or the core network function.

[0357] In Figure 5 's example, the first device is the access network device, the second device is the core network function, the third device is the core network function, and the first model is deployed on the first device.

[0358] Such as Figure 5 shown, the inference method may include the following steps:

[0359] S503, the access network device sends second information to the core network function.

[0360] Among them, the second information includes the requirement information for reasoning with the first model and / or the auxiliary information for determining the first information.

[0361] S504, the core network function sends the first information to the access network device. Among them, the first information is determined according to the second information.

[0362] Among them, the first information includes at least one of prompt indication information, knowledge indication information, and fine-tuning dataset indication information.

[0363] In some embodiments, before S503, the method further includes:

[0364] S501, the access network device may send third information to the core network function;

[0365] S502, the access network device receives the first model sent by the core network function.

[0366] In summary, in the embodiments of the present application, the first device (i.e., the inference node) may obtain the auxiliary information for reasoning with the first model from the auxiliary inference node, such as prompt indication information, knowledge indication information, fine-tuning dataset, etc. Thus, the inference node may use the first model for reasoning based on the prompt indication information, knowledge indication information, fine-tuning dataset, etc., which is beneficial to improving the inference performance of the model.

[0367] As described above in conjunction with Figures 2 to 5 , the method embodiments of the present application are described in detail. Below in conjunction with Figures 6 to 11, the device embodiments of the present application are described in detail. It should be understood that the device embodiments correspond to the method embodiments, and similar descriptions can refer to the method embodiments.

[0368] In the inference method based on the AI model in the communication system provided by the embodiments of the present application, the execution subject can be a communication device. In the embodiments of the present application, taking the communication device in the communication system executing the inference method based on the AI model in the communication system as an example, the communication device provided by the embodiments of the present application is described.

[0369] Figure 6 FIG. shows a schematic block diagram of a communication device 600 according to an embodiment of the present application. As Figure 6 shown, the communication device 600 includes:

[0370] A processing unit 610, configured to obtain first information, wherein a first model is deployed on the communication device 600;

[0371] According to the first information, adapt the first model to a target task;

[0372] Wherein, the first model is a compressed AI model, and the first information includes at least one of the following information:

[0373] Prompt indication information, knowledge indication information, fine-tuning dataset indication information.

[0374] In some embodiments, the prompt indication information includes at least one of the following:

[0375] A thought chain related to the target task;

[0376] Prompt text related to the target task;

[0377] A prompt file related to the target task;

[0378] Prompt code related to the target task;

[0379] A thought chain identifier related to the target task;

[0380] A prompt text identifier related to the target task;

[0381] A prompt file identifier related to the target task;

[0382] A prompt code identifier related to the target task;

[0383] A prompt file address related to the target task.

[0384] In some embodiments, the knowledge indication information, the knowledge indication information includes at least one of knowledge graph indication information and knowledge vector library indication information;

[0385] Among them, the knowledge graph indication information includes at least one of the following:

[0386] The relationships between entities in the knowledge graph;

[0387] The attribute information of entities in the knowledge graph;

[0388] The data structure of the knowledge graph;

[0389] The correlation between entities in the knowledge graph;

[0390] The subject-predicate-object (SPO) triples of the statements in the knowledge graph;

[0391] The knowledge graph file;

[0392] The knowledge graph identifier;

[0393] The knowledge graph file address.

[0394] In some embodiments, the fine-tuning dataset indication information is used to indicate the input information and label information of the first model, and the input information of the first model is related to the target task.

[0395] In some embodiments, the input information of the first model includes at least one of the following:

[0396] Measurement results of wireless signals, sensing results, wireless statistical characteristics.

[0397] In some embodiments, the sensing result is obtained by sensing a wireless signal or by a third-party device, where the third-party device includes at least one of the following:

[0398] Camera, radar, lidar, Global Positioning System (GPS), Inertial Measurement Unit.

[0399] In some embodiments, the wireless statistical characteristics include at least one of the following:

[0400] Occlusion probability, user state, user behavior, handover failure rate.

[0401] In some embodiments, the user state includes at least one of the following:

[0402] Registration management state, connection management state, Radio Resource Control (RRC) connection management state.

[0403] In some embodiments, the user behavior includes at least one of the following:

[0404] Behaviors in the RRC idle state, behaviors in the RRC inactive state, behaviors in the RRC connected state.

[0405] In some embodiments, the communication device further includes

[0406] a communication unit, configured to receive the first information from a second device;

[0407] wherein, the communication device 600 is a terminal, and the second device is an access network device or a core network function; or

[0408] the communication device 600 is an access network device, and the second device is a core network function.

[0409] In some embodiments, the communication device 600 is a terminal, and the second device is an access network device, and the first information is carried in at least one of the following signaling: layer 1 signaling, layer 2 signaling, layer 3 signaling, AI layer signaling, data plane signaling; or

[0410] the communication device 600 is a terminal, and the second device is a core network function, and the first information is carried in at least one of the following signaling: non-access stratum NAS signaling, AI layer signaling, data plane signaling.

[0411] In some embodiments, the communication device 600 is a terminal, and the communication device further includes

[0412] a communication unit, configured to send second information to the second device, where the second information includes at least one of the following:

[0413] requirement information for inference by the first model;

[0414] auxiliary information for determining the first information;

[0415] identification information of at least one event for triggering the first device to send the requirement information or the auxiliary information.

[0416] In some embodiments, the first information is determined based on the second information.

[0417] In some embodiments, the auxiliary information for determining the first information includes at least one of the following:

[0418] identification information of the AI model, function indication of the AI model, characteristic indication of the AI model, configuration indication of the communication device 600.

[0419] In some embodiments, the requirement information for inference by the first model includes at least one of the following:

[0420] target task indication for adaptation required by the AI model;

[0421] Quality of Experience (QoE) requirements for the AI model;

[0422] Business processing latency of the AI model;

[0423] Business processing accuracy of the AI model;

[0424] Business computing volume of the AI model;

[0425] Data processing scale of the AI model.

[0426] In some embodiments, the communication device 600 is a terminal, the second device is an access network device, and the second information is carried in at least one of the following signaling: Layer 1 signaling, Layer 2 signaling, Layer 3 signaling, AI layer signaling, data plane signaling; or

[0427] The communication device 600 is a terminal, the second device is a core network function, and the second information is carried in at least one of the following signaling: NAS signaling, AI layer signaling, data plane signaling.

[0428] In some embodiments, the communication device 600 is a terminal, and the communication device further includes:

[0429] A communication unit, configured to send the second information to the second device when at least one of the following events occurs:

[0430] Receiving a first indication for indicating activation or switching of the model;

[0431] The communication device 600 determines to activate or switch the model;

[0432] The communication device 600 determines that the currently used model does not meet the requirements;

[0433] Receiving a second indication for indicating that the currently used model does not meet the requirements.

[0434] In some embodiments, the communication device further includes

[0435] A communication unit, configured to receive the first model sent by a third device;

[0436] Wherein, the communication device 600 is a terminal, and the third device is an access network device or a core network function or a third-party external server; or

[0437] The communication device 600 is an access network device, and the third device is a core network function or a third-party external server.

[0438] In some embodiments, the communication device further includes:

[0439] A communication unit, configured to send third information to the third device, where the first model is selected based on the third information;

[0440] Wherein, the third information includes at least one of the following:

[0441] Identification information of the AI model, function indication of the AI model, characteristic indication of the AI model, configuration indication of the communication device 600.

[0442] In some embodiments, the communication device 600 is a terminal, the third device is an access network device, and the third information or the first model is carried in at least one of the following signaling: layer 1 signaling, layer 2 signaling, layer 3 signaling, data plane signaling, AI layer signaling; or

[0443] The communication device 600 is a terminal, the third device is a core network function, and the third information or the first model is carried in at least one of the following signaling: NAS signaling, data plane signaling, AI layer signaling.

[0444] In some embodiments, the configuration indication of the communication device 600 is used to indicate at least one of the following:

[0445] Number of transmit antennas, number of transmit beams, number of antenna ports, transmit power, antenna gain, 3dB bandwidth of the beam, spacing, frequency, system bandwidth.

[0446] In some embodiments, the communication device 600 is a terminal, and the communication device further includes:

[0447] A communication unit, configured to send at least one of fourth information and first capability information to a network-side device, where the fourth information is used to indicate auxiliary information required for the terminal to perform inference based on the first model, and the first capability information is used to indicate the capability of the terminal to perform inference based on the first model.

[0448] In some embodiments, the fourth information includes at least one of the following:

[0449] Indication information of a prompt for adapting to the target task;

[0450] Indication information of knowledge for adapting to the target task;

[0451] Indication information of a fine-tuning data set for adapting to the target task.

[0452] In some embodiments, the first capability information includes at least one of the following:

[0453] Identification information of the first model supported by the terminal;

[0454] Whether the terminal supports adapting the first model to the target task based on the prompt information;

[0455] Whether the terminal supports adapting the first model to the target task based on the fine-tuning data set;

[0456] Whether the terminal supports adapting the first model to the target task based on the knowledge information;

[0457] The indication information of the prompt supported by the terminal;

[0458] The indication information of the knowledge supported by the terminal;

[0459] The indication information of the fine-tuning data set supported by the terminal.

[0460] In some embodiments, the communication device 600 is a terminal, and the communication device further includes:

[0461] A communication unit, configured to send at least one of the fourth information and the first capability information to the network-side device during a model registration process or a capability reporting process.

[0462] Optionally, in some embodiments, the above communication unit may be a communication interface or a transceiver, or an input / output interface of a communication chip or a system-on-chip. The above processing unit may be one or more processors.

[0463] It should be understood that the communication device 600 according to the embodiments of the present application may correspond to the first device in the method embodiments of the present application, and the above and other operations and / or functions of each unit in the communication device 600 are respectively for implementing Figures 2 to 5 the corresponding processes of the first device in the method embodiments shown, and achieve the same technical effects. To avoid repetition, details are not described here again.

[0464] Figure 7 Fig. shows a schematic block diagram of a communication device 700 according to an embodiment of the present application. As Figure 7 shown, the communication device 700 includes:

[0465] A communication unit 710, configured to send first information to a first device, where a first model is deployed on the first device, and the first information is used for the first device to adapt the first model to a target task;

[0466] Wherein, the first information includes at least one of the following information:

[0467] Prompt indication information, knowledge indication information, fine-tuning data set indication information.

[0468] In some embodiments, the prompt indication information includes at least one of the following:

[0469] A chain of thought related to the target task;

[0470] Prompt text related to the target task;

[0471] Prompt file related to the target task;

[0472] Prompt code related to the target task;

[0473] Chain of thought identifier related to the target task;

[0474] Prompt text identifier related to the target task;

[0475] Prompt file identifier related to the target task;

[0476] Prompt code identifier related to the target task;

[0477] Prompt file address related to the target task.

[0478] In some embodiments, the knowledge indication information includes at least one of knowledge graph indication information and knowledge vector library indication information;

[0479] Wherein, the knowledge graph indication information includes at least one of the following:

[0480] Relationships between entities in the knowledge graph;

[0481] Attribute information of entities in the knowledge graph;

[0482] Data structure of the knowledge graph;

[0483] Correlations between entities in the knowledge graph;

[0484] Subject-property-object SPO triples of statements in the knowledge graph;

[0485] Knowledge graph file;

[0486] Knowledge graph identifier;

[0487] Knowledge graph file address.

[0488] In some embodiments, the fine-tuning dataset indication information is used to indicate the input information and label information of the first model, and the input information of the first model is related to the target task.

[0489] In some embodiments, the input information of the first model includes at least one of the following:

[0490] Measurement results, perception results, and wireless statistical characteristics of wireless signals.

[0491] In some embodiments, the perception result is obtained by perceiving a wireless signal, or obtained by a third-party device, where the third-party device includes at least one of the following:

[0492] Camera, radar, lidar, Global Positioning System (GPS), Inertial Measurement Unit.

[0493] In some embodiments, the wireless statistical characteristics include at least one of the following:

[0494] Occlusion probability, user status, user behavior, handover failure rate.

[0495] In some embodiments, the user status includes at least one of the following:

[0496] Registration management status, connection management status, Radio Resource Control (RRC) connection management status.

[0497] In some embodiments, the user behavior includes at least one of the following:

[0498] Behavior in the RRC idle state, behavior in the RRC inactive state, behavior in the RRC connected state.

[0499] In some embodiments, the first device is a terminal, the communication device 700 is an access network device, and the first information is carried in at least one of the following signaling: Layer 1 signaling, Layer 2 signaling, Layer 3 signaling, AI layer signaling, data plane signaling; or

[0500] The first device is a terminal, the communication device 700 is a core network function, and the first information is carried in at least one of the following signaling: Non-Access Stratum (NAS) signaling, AI layer signaling, data plane signaling.

[0501] In some embodiments, when the first device is a terminal, the communication unit is further configured to:

[0502] Receive second information sent by the first device, where the second information includes at least one of the following:

[0503] Requirement information for inference by the first model;

[0504] Auxiliary information for determining the first information;

[0505] Identification information of at least one event that triggers the first device to send the requirement information or the auxiliary information.

[0506] In some embodiments, the first information is determined based on the second information.

[0507] In some embodiments, the auxiliary information for determining the first information includes at least one of the following:

[0508] The identification information of the AI model, the function indication of the AI model, the characteristic indication of the AI model, and the configuration indication of the first device.

[0509] In some embodiments, the requirement information for the first model to perform inference includes at least one of the following:

[0510] The target task indication that the AI model needs to adapt to;

[0511] The quality of experience QoE requirement of the AI model for the service;

[0512] The service processing delay of the AI model;

[0513] The service processing accuracy of the AI model;

[0514] The service calculation amount of the AI model;

[0515] The data processing scale of the AI model.

[0516] In some embodiments, the first device is a terminal, the communication device 700 is an access network device, and the second information is carried in at least one of the following signaling: layer 1 signaling, layer 2 signaling, layer 3 signaling, AI layer signaling, data plane signaling; or

[0517] The first device is a terminal, the communication device 700 is a core network function, and the second information is carried in at least one of the following signaling: NAS signaling, AI layer signaling, data plane signaling.

[0518] Optionally, in some embodiments, the above communication unit may be a communication interface or a transceiver, or an input / output interface of a communication chip or a system-on-chip.

[0519] It should be understood that the communication device 700 for signal forwarding according to the embodiments of the present application may correspond to the second device in the method embodiments of the present application, and the above and other operations and / or functions of each unit in the communication device 700 are respectively for implementing Figures 2 to 5 the corresponding processes of the second device in the method embodiments shown in, and achieving the same technical effects. To avoid repetition, it will not be elaborated here.

[0520] In some embodiments, the apparatuses 600 and 700 in the embodiments of the present application may be electronic devices, such as electronic devices with an operating system, or components in an electronic device, such as an integrated circuit or a chip. The electronic device may be a terminal or other devices other than a terminal. Exemplarily, the terminal may include, but is not limited to, the types of the terminal 11 listed above, and other devices may be a server, a Network Attached Storage (NAS), etc., which are not specifically limited in the embodiments of the present application.

[0521] As Figure 8 shown, the embodiments of the present application further provide a communication device 1000, including a processor 1001 and a memory 1002. A program or instruction that can run on the processor 1001 is stored on the memory 1002. For example, when the communication device 1000 is a first device, the steps performed by the first device in the above-mentioned inference method embodiments are implemented when the program or instruction is executed by the processor 1001, and the same technical effects can be achieved. For example, when the communication device 1000 is a second device, the steps performed by the second device in the above-mentioned inference method embodiments are implemented when the program or instruction is executed by the processor 1001, and the same technical effects can be achieved. For example, when the communication device 1000 is a third device, the steps performed by the third device in the above-mentioned inference method embodiments are implemented when the program or instruction is executed by the processor 1001, and the same technical effects can be achieved. To avoid repetition, details are not described here again.

[0522] The embodiments of the present application further provide a terminal, including a processor and a communication interface. The communication interface is coupled to the processor. The processor is configured to run a program or instruction to implement the steps in the method embodiment as Figures 2 to 5 shown. This terminal embodiment corresponds to the above-mentioned terminal-side method embodiment. Each implementation process and implementation manner of the above method embodiment can be applied to this terminal embodiment, and the same technical effects can be achieved. Specifically, Figure 9 FIG. is a schematic hardware structure diagram of a terminal for implementing the embodiments of the present application.

[0523] The terminal 1100 includes, but is not limited to, at least some components such as a radio frequency unit 1101, a network module 1102, an audio output unit 1103, an input unit 1104, a sensor 1105, a display unit 1106, a user input unit 1107, an interface unit 1108, a memory 1109, and a processor 1110.

[0524] Those skilled in the art can understand that the terminal 1100 may further include a power source (such as a battery) for powering each component. The power source can be logically connected to the processor 1110 through a power management system, so as to manage functions such as charging, discharging, and power consumption management through the power management system. Figure 9 The terminal structure shown in Figure 9 does not limit the terminal. The terminal may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements, which will not be elaborated here.

[0525] It should be understood that in the embodiments of the present application, the input unit 1104 may include a Graphics Processing Unit (GPU) 11041 and a microphone 11042. The graphics processor 11041 processes the image data of still pictures or videos obtained by an image capture device (such as a camera) in a video capture mode or an image capture mode. The display unit 1106 may include a display panel 11061, and the display panel 11061 can be configured in the form of a liquid crystal display, an organic light-emitting diode, etc. The user input unit 1107 includes at least one of a touch panel 11071 and other input devices 11072. The touch panel 11071 is also called a touch screen. The touch panel 11071 may include two parts: a touch detection device and a touch controller. The other input devices 11072 may include, but are not limited to, a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, and a joystick, which will not be elaborated here.

[0526] In the embodiments of the present application, after receiving downlink data from a network-side device, the radio frequency unit 1101 can transmit it to the processor 1110 for processing; in addition, the radio frequency unit 1101 can send uplink data to the network-side device. Generally, the radio frequency unit 1101 includes, but is not limited to, an antenna, an amplifier, a transceiver, a coupler, a low-noise amplifier, a duplexer, etc.

[0527] The memory 1109 can be used to store software programs or instructions as well as various data. The memory 1109 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data. Among them, the first storage area may store an operating system, application programs or instructions required for at least one function (such as a sound playback function, an image playback function, etc.). In addition, the memory 1109 may include a volatile memory or a non-volatile memory. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate synchronous dynamic random access memory (DDR SDRAM), an enhanced synchronous dynamic random access memory (ESDRAM), a synch link dynamic random access memory (SLDRAM), and a direct rambus random access memory (DRRAM). The memory 1109 in the embodiments of the present application includes, but is not limited to, these and any other suitable types of memories.

[0528] The processor 1110 may include one or more processing units; optionally, the processor 1110 integrates an application processor and a modem processor. Among them, the application processor mainly processes operations related to the operating system, user interface, and application programs, etc., and the modem processor mainly processes wireless communication signals, such as a baseband processor. It can be understood that the above-mentioned modem processor may not be integrated into the processor 1110 either.

[0529] It can be understood that the implementation processes of the various implementation manners mentioned in this embodiment can refer to Figures 2 to 5 the relevant descriptions of the method embodiments shown, and achieve the same or corresponding technical effects. To avoid repetition, they will not be elaborated here.

[0530] The embodiments of the present application further provide a network-side device, including a processor and a communication interface. The communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement as Figures 2 to 5The steps of the method embodiments shown. This network-side device embodiment corresponds to the above-mentioned access network device side or core network function side method embodiments. Each implementation process and realization method of the above method embodiments can be applied to this network-side device embodiment, and the same technical effects can be achieved.

[0531] Specifically, an embodiment of the present application further provides a network-side device. As Figure 10 shown, the network-side device 1200 includes: an antenna 1201, a radio frequency device 1202, a baseband device 1203, a processor 1204, and a memory 1205. The antenna 1201 is connected to the radio frequency device 1202. In the uplink direction, the radio frequency device 1202 receives information through the antenna 1201 and sends the received information to the baseband device 1203 for processing. In the downlink direction, the baseband device 1203 processes the information to be sent and sends it to the radio frequency device 1202. After processing the received information, the radio frequency device 1202 sends it out through the antenna 1201.

[0532] The method executed by the network-side device in the above embodiments can be implemented in the baseband device 1203, and the baseband device 1203 includes a baseband processor.

[0533] The baseband device 1203 may include, for example, at least one baseband board. A plurality of chips are provided on the baseband board. As shown in FIG. 120, one of the chips is, for example, a baseband processor, which is connected to the memory 1205 through a bus interface to call the program in the memory 1205 and execute the network device operations shown in the above method embodiments.

[0534] The network-side device may further include a network interface 1206, and this interface is, for example, a Common Public Radio Interface (CPRI).

[0535] Specifically, the network-side device 1200 in the embodiment of the present application further includes: instructions or programs stored on the memory 1205 and executable on the processor 1204. The processor 1204 calls the instructions or programs in the memory 1205 to execute Figures 6 to 7 the methods executed by the modules shown and achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0536] Specifically, an embodiment of the present application further provides a network-side device. As Figure 11 shown, the network-side device 1300 includes: a processor 1301, a network interface 1302, and a memory 1303. Among them, the network interface 1302 is, for example, a Common Public Radio Interface (CPRI).

[0537] Specifically, the network-side device 1300 according to an embodiment of the present invention further includes: instructions or programs stored in the memory 1303 and executable on the processor 1301. The processor 1301 invokes the instructions or programs in the memory 1303 to execute Figures 6 to 7 the methods executed by the modules shown in the figure, and achieve the same technical effects. To avoid repetition, they will not be elaborated here.

[0538] An embodiment of the present application further provides a readable storage medium. Programs or instructions are stored on the readable storage medium. When the programs or instructions are executed by a processor, each process of the above-mentioned embodiment of the inference method based on the AI model in the communication system is implemented, and the same technical effects can be achieved. To avoid repetition, they will not be elaborated here.

[0539] Wherein, the processor is the processor in the communication device, communication equipment, terminal or network-side device described in the above embodiment. The readable storage medium includes computer-readable storage media, such as computer read-only memory ROM, random access memory RAM, magnetic disks or optical discs, etc. In some examples, the readable storage medium may be a non-transitory readable storage medium.

[0540] Another embodiment of the present application provides a chip. The chip includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement each process of the above-mentioned embodiment of the inference method based on the AI model in the communication system, and the same technical effects can be achieved. To avoid repetition, they will not be elaborated here.

[0541] It should be understood that the chip mentioned in the embodiment of the present application may also be referred to as a system-on-chip, system chip, chip system or system-on-chip, etc.

[0542] Another embodiment of the present application provides a computer program / program product. The computer program / program product is stored in a storage medium. The computer program / program product is executed by at least one processor to implement each process of the above-mentioned embodiment of the inference method based on the AI model in the communication system, and the same technical effects can be achieved. To avoid repetition, they will not be elaborated here.

[0543] An embodiment of the present application further provides a communication system, including: a first device, a second device and a third device. The first device can be used to execute the steps executed by the first device in the above-mentioned inference method based on the AI model in the communication system. The second device can be used to execute the steps executed by the second device in the above-mentioned inference method based on the AI model in the communication system. The third device can be used to execute the steps executed by the third device in the above-mentioned inference method based on the AI model in the communication system.

[0544] It should be noted that in this text, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising a..." does not exclude the existence of additional identical elements in the process, method, article or device comprising such element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, but may also include performing functions in a substantially simultaneous manner or in the reverse order according to the functions involved. For example, the described methods may be performed in an order different from that described, and various steps may be added, omitted or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0545] Through the description of the above embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of a computer software product plus a necessary general hardware platform, and of course, can also be implemented by hardware. The computer software product is stored in a storage medium (such as ROM, RAM, magnetic disk, optical disk, etc.) and includes several instructions for causing a terminal or a network-side device to execute the methods described in various embodiments of the present application.

[0546] The embodiments of the present application have been described above in conjunction with the accompanying drawings. However, the present application is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present application, those of ordinary skill in the art can also make many forms of embodiments without departing from the purpose of the present application and the scope protected by the claims. These embodiments are all within the protection scope of the present application.

Claims

1. A reasoning method based on an artificial intelligence (AI) model, characterized in that: include: A first device acquires first information associated with a target task, wherein a first model is deployed on the first device; Adapting the first model to a target task according to the first information; The first information includes at least one of the following information: prompt indication information, knowledge indication information, and fine-tuning data set indication information.

2. The method according to claim 1, characterized in that: The prompt indication information includes at least one of the following: The chain of thoughts related to the stated goal task; Prompt text related to the target task; Prompt files related to the target task; a prompt code associated with the target task; Identification of thought chains related to the target task; Prompt text labels related to the target task; The reminder file identifier related to the target task; A prompt code identifier related to the target task; The address of the prompt file related to the target task.

3. The method according to claim 1 or 2, characterized in that: The knowledge indication information includes at least one of knowledge graph indication information and knowledge vector library indication information; The knowledge graph indication information includes at least one of the following: The relationships between entities in the knowledge graph; Attribute information of entities in the knowledge graph; The data structure of the knowledge graph; The correlation between entities in the knowledge graph; The subject attribute object SPO triple of the sentence in the knowledge graph; Knowledge graph files; Knowledge graph identification; Knowledge graph file address.

4. The method according to any one of claims 1 to 3, characterized in that The fine-tuning dataset indication information is used to indicate input information and label information of the first model, where the input information of the first model is related to the target task; The input information of the first model includes at least one of the following: Measurement results, perception results, and wireless statistical characteristics of wireless signals; The sensing result is obtained by sensing the wireless signal, or by a third-party device, wherein the third-party device includes at least one of the following: Cameras, radars, lidar, global positioning system GPS, inertial measurement units; The wireless statistical characteristics include at least one of the following: Occlusion probability, user status, user behavior, switching failure rate; The user status includes at least one of the following: Registration management status, connection management status, radio resource control RRC connection management status; The user behavior includes at least one of the following: Behavior in RRC idle state, behavior in RRC inactive state, and behavior in RRC connected state.

5. The method according to any one of claims 1 to 4, characterized in that The first device acquires first information, including: The first device receives the first information from the second device; The first device is a terminal, and the second device is an access network device or a core network function; or The first device is an access network device, and the second device is a core network function.

6. The method according to claim 5, characterized in that The method further comprises: The first device sends second information to the second device, where the second information includes at least one of the following: Requirement information for performing reasoning on the first model; Auxiliary information used to determine the first information; identification information of at least one event that triggers the first device to send the requirement information or the auxiliary information; The auxiliary information used to determine the first information includes at least one of the following: Identification information of the AI ​​model, a function indication of the AI ​​model, a characteristic indication of the AI ​​model, and a configuration indication of the first device; The requirement information used for the first model to perform reasoning includes at least one of the following: Indication of target tasks that the AI ​​model needs to adapt to; Service quality of experience (QoE) requirements for AI models; Business processing latency of AI models; The accuracy of AI model’s business processing; The business computing volume of AI models; The data processing scale of AI models.

7. The method according to claim 6, characterized in that The first device is a terminal, and the first device sends second information to the second device, including: When at least one of the following events occurs, sending the second information to the second device: A first indication is received, where the first indication is used to instruct activation or switching of a model; The first device determines to activate or switch a model; The first device determines that the currently used model does not meet the requirements; A second indication is received, where the second indication is used to indicate that the currently used model does not meet the requirement.

8. The method according to claim 6, characterized in that The configuration indication of the first device is used to indicate at least one of the following: Number of transmit antennas, number of transmit beams, number of antenna ports, transmit power, antenna gain, beam 3dB bandwidth, spacing, frequency, and system bandwidth.

9. The method according to any one of claims 1 to 8, characterized in that When the first device is a terminal, the method further includes: The terminal sends at least one of fourth information and first capability information to the network side device, wherein the fourth information is used to indicate auxiliary information required for the terminal to perform reasoning based on the first model, and the first capability information is used to indicate the ability of the terminal to perform reasoning based on the first model.

10. The method according to claim 9, characterized in that The fourth information includes at least one of the following: Indicative information for adapting prompts for the target task; Indicative information for adapting the knowledge of the target task; Indicative information of a fine-tuning dataset adapted to the target task.

11. The method according to claim 9 or 10, characterized in that: The first capability information includes at least one of the following: identification information of the first model supported by the terminal; Whether the terminal supports adapting the first model to the target task based on the prompt information; Whether the terminal supports adapting the first model to the target task based on a fine-tuning dataset; whether the terminal supports adapting the first model to the target task based on knowledge information; Indicative information of prompts supported by the terminal; Indicative information of knowledge supported by the terminal; Indicative information of the fine-tuning datasets supported by the terminal.

12. The method according to any one of claims 9 to 11, characterized in that: The terminal sends at least one of the fourth information and the first capability information to the network side device, including: The terminal sends at least one of the fourth information and the first capability information to the network side device during a model registration process or a capability reporting process.

13. A reasoning method based on an artificial intelligence (AI) model, characterized in that: include: The second device sends first information to the first device, the first model is deployed on the first device, and the first information is used by the first device to adapt the first model to the target task; The first information includes at least one of the following information: Hint instructions, knowledge instructions, fine-tuning dataset instructions.

14. A communication device, characterized in that: include: a processing unit, configured to obtain first information, wherein a first model is deployed on the communication device; Adapting the first model to a target task according to the first information; The first information includes at least one of the following information: prompt indication information, knowledge indication information, and fine-tuning data set indication information.

15. A communication device, characterized in that: include: A communication unit, configured to send first information to a first device, on which a first model is deployed, and wherein the first information is used by the first device to adapt the first model to a target task; The first information includes at least one of the following information: Hint instructions, knowledge instructions, fine-tuning dataset instructions.

16. A communication device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 12 or the steps of the method according to claim 13 are implemented.

17. A readable storage medium, characterized in that: The readable storage medium stores a program or an instruction, and when the program or the instruction is executed by a processor, the steps in the method according to any one of claims 1 to 12 or the steps in the method according to claim 13 are implemented.

Images