DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Claim Objections
Claims 14 and 16 are objected to because of the following informalities:
Regarding claim 14 – Line 2, “toto” needs to be amended to -- to --.
Regarding claim 16 - Line 4, “toto” needs to be amended to -- to --.
Appropriate correction is required.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-10, 12, 14-24, 26, 31, and 32 are rejected under 35 U.S.C. 103 as being unpatentable over Jin et al. (US 2025/0317254 A1) hereinafter Jin in view of Sajid et al. (US 2026/0099447 A1) hereinafter Sajid.
Regarding claim 1 – Jin discloses A user equipment (UE) for wireless communication, the UE comprising: refer to Figure 1, also paragraph [0156] - The model training node may be a terminal device, a network device, or another network element having an AI function in a communication system.
one or more memories; and one or more processors, coupled to the one or more memories, the one or more processors configured to cause the UE to: refer to paragraph [0090] - a communication apparatus is provided, including a processor and a storage medium. The storage medium stores instructions.
generate an input token sequence associated with one or more of: wireless communication data between the UE and a network unit, configuration data associated with the UE, or configuration data associated with the network unit; refer to paragraph [0010] - the channel information may be predicted in a plurality of manners. For example, the channel information may be predicted in one or more of the following manners: linear fitting, Kalman filtering, an AI model, or the like. An AI model-based prediction manner is used as an example. Channel information measured at one or more historical moments is input into an AI prediction model, and the AI prediction model outputs channel information at one or more future moments, also, refer to paragraph [0122] - To support the AI technology in the wireless network, a dedicated AI network element or AI module may be further introduced into the network. If the AI network element is introduced, the AI network element is corresponding to an independent network element. If the AI module is introduced, the AI module may be located in a specific network element, and the corresponding network element may be a terminal device, a network device, or the like, also, refer to paragraph [0125] - the network device 110 may send data related to training of the AI model to the AI network element 140, and the AI network element 140 constructs a training dataset and trains the AI model. For example, the data related to training of the AI model may include data reported by the terminal device. The AI network element 140 may send a result of an operation related to the AI model to the network device 110, and the network device 110 forwards the result of the operation related to the AI model to the terminal device. For example, the result of the operation related to the AI model may include at least one of the following: a trained AI model, an evaluation result or a test result for the model, and the like. For example, a part of the trained AI model may be deployed on the network device 110, and the other part is deployed on the terminal device. Alternatively, the trained AI model may be deployed on the network device 110. Alternatively, the trained AI model may be deployed on the terminal device. (The examiner interprets the training dataset as the input token sequence.)
Jin does not explicitly disclose obtain one or more predicted next tokens representing predicted wireless communication data from one or more generative artificial intelligence (AI) based language models using the input token sequence, wherein the one or more generative AI based language models are pretrained with a dataset comprising data token sequences associated with wireless protocol data.
Sajid discloses obtain one or more predicted next tokens representing predicted wireless communication data from one or more generative artificial intelligence (AI) based language models using the input token sequence, wherein the one or more generative AI based language models are pretrained with a dataset comprising data token sequences associated with wireless protocol data, refer to paragraph [0020] - A token computed based on a user input to a generative large language model may be received. A portion of the generative large language model may be identified using the token and a machine learning model trained to identify portions of the generative large language model. The portion may be written into a memory. The generative large language model may be caused to generate an output based on the user input using the portion in the memory, also, paragraph [0021] - A current token and at least one previous token generated based on a user input to a generative large language model may be received. Subnetworks within the generative large language model may be identified by processing the current token and the at least one previous token using a machine learning model. The subnetworks may be prefetched from a first memory into a second memory. The generative large language model may be caused to generate an output based on the user input using the subnetworks in the second memory, also, refer to Figure 9 and paragraph [0076] - FIG. 9 illustrates a representation of determining potential subnetworks within layers of a generative large language model 160 based on tokens, according to embodiments of the disclosure. The example shown in FIG. 9 includes a first table 902, a second table 904, and the generative large language model 160. The first table 902 is illustrated to include tokens (e.g., all of the tokens) in the vocabulary of the generative large language model 160, and claim 1.
It would have been obvious to an ordinary person of skill in the art prior to the effective date of the application to combine Jin with Sajid to provide Jin with obtain one or more predicted next tokens representing predicted wireless communication data from one or more generative artificial intelligence (AI) based language models using the input token sequence, wherein the one or more generative AI based language models are pretrained with a dataset comprising data token sequences associated with wireless protocol data.
The suggestion/motivation would have been that Jin discloses – refer to paragraph [0005] - However, in an actual scenario, predicted CSI possibly cannot accurately reflect actual channel quality, affecting accuracy of a subsequent related configuration of the network device. For example, when the terminal device feeds back CSI, if an artificial intelligence (artificial intelligence, AI) prediction model is deployed on the terminal device, the terminal device may feed back, to the network device, CSI predicted by the AI prediction model.
The benefit being the AI prediction model has the potential to be more accurate – refer to paragraph [0010] - An AI model-based prediction manner is used as an example. Channel information measured at one or more historical moments is input into an AI prediction model, and the AI prediction model outputs channel information at one or more future moments.
Jin discloses generate one or more communication parameters for future wireless communication by the UE, based at least in part on the one or more predicted next tokens, refer to paragraph [0010] - An AI model-based prediction manner is used as an example. Channel information measured at one or more historical moments is input into an AI prediction model, and the AI prediction model outputs channel information at one or more future moments.
Regarding claim 2 – Jin and Sajid disclose claim 1. Jin discloses wherein the one or more generative AI based language models are implemented at one or more of: the UE, a server in communication with the UE, or the network unit, refer to paragraph [0122] - To support the AI technology in the wireless network, a dedicated AI network element or AI module may be further introduced into the network. If the AI network element is introduced, the AI network element is corresponding to an independent network element. If the AI module is introduced, the AI module may be located in a specific network element, and the corresponding network element may be a terminal device, a network device, or the like, also, refer to Figure 2 and paragraph [0125] - In a possible implementation, the network device 110 may send data related to training of the AI model to the AI network element 140, and the AI network element 140 constructs a training dataset and trains the AI model. For example, the data related to training of the AI model may include data reported by the terminal device. The AI network element 140 may send a result of an operation related to the AI model to the network device 110, and the network device 110 forwards the result of the operation related to the AI model to the terminal device. For example, the result of the operation related to the AI model may include at least one of the following: a trained AI model, an evaluation result or a test result for the model, and the like. For example, a part of the trained AI model may be deployed on the network device 110, and the other part is deployed on the terminal device. Alternatively, the trained AI model may be deployed on the network device 110. Alternatively, the trained AI model may be deployed on the terminal device.
Regarding claim 3 – Jin and Sajid disclose claim 1. Jin discloses wherein the wireless communication data between the UE and the network unit comprises one or more of: UE-side wireless communication data that is observed by the UE at a current communication slot; network-side wireless communication data that is signaled to the UE at the current communication slot; previous wireless communication data from previous communication slots retrieved from a memory; or simulated or synthetic wireless data, refer to Figure 8 and paragraph [378] - The terminal device may predict CSI at one future moment based on the m pieces of measured CSI. Specifically, the terminal device may predict the CSI at the future moment by using a prediction model. To be specific, the m pieces of measured CSI are input into the prediction model for processing, to predict the CSI at the future moment.
Regarding claim 4 – Jin and Sajid disclose claim 1. Jin discloses wherein the wireless communication data between the UE and the network unit comprises one or more of: a channel quality indicator (CQI) report; a decoded grant from the network unit, the decoded grant indicative of one or more of: a rank, a spectral efficiency, or a size of the grant; a UE hybrid automatic repeat request (HARQ) per grant; a raw channel measurement; a buffer status at the UE; a scheduling request; or an absence of all activity, refer to Figure 9 and paragraph [0389] - The terminal device may measure CSI based on the CSI-RS, to obtain CSI measured at the moment t1 (an example of channel information measured based on the reference signal #1). The moment t1′ is earlier than the moment t1. The CSI-RS is a CSI-RS shown by a solid arrow in FIG. 9.
Regarding claim 5 – Jin and Sajid disclose claim 1. Jin discloses wherein the predicted wireless communication data comprises synthetic configuration data associated with the network unit, refer to paragraph [0125] - In a possible implementation, the network device 110 may send data related to training of the AI model to the AI network element 140, and the AI network element 140 constructs a training dataset and trains the AI model. For example, the data related to training of the AI model may include data reported by the terminal device. The AI network element 140 may send a result of an operation related to the AI model to the network device 110, and the network device 110 forwards the result of the operation related to the AI model to the terminal device. For example, the result of the operation related to the AI model may include at least one of the following: a trained AI model, an evaluation result or a test result for the model, and the like. For example, a part of the trained AI model may be deployed on the network device 110, and the other part is deployed on the terminal device. Alternatively, the trained AI model may be deployed on the network device 110, also, refer to paragraph [0136] - The auto-encoder is an unsupervised learning neural network, and is characterized by using input data as label data. Therefore, the auto-encoder may also be understood as a self-supervised learning neural network. The auto-encoder may be configured to compress and restore data. For example, the encoder in the auto-encoder may compress (encode) data A to obtain data B, and the decoder in the auto-encoder may decompress (decode) the data B to restore the data A, also, refer to paragraph [0144] - The training dataset is used to train an AI model. The training dataset may include an input of the AI model, or include the input of the AI model and a target output of the AI model. The training dataset includes one or more pieces of training data. The training data may include a training sample input to the AI model, or may include the target output of the AI model. The target output may also be referred to as a label or a label sample, also, refer to paragraph [0145] - In the communication field, the training dataset may include simulation data collected by using a simulation platform, or may include experimental data collected in an experiment scenario, or may include actual measurement data collected in an actual communication network, also, refer to paragraph [0152] -Training processes for different models may be deployed on different devices or nodes, or may be deployed on a same device or node. Inference processes for different models may be deployed on different devices or nodes, or may be deployed on a same device or node. For example, a model training phase is performed by a terminal device. After training an encoder and a decoder that match each other, the terminal device may send a model parameter of the decoder to a network device. For example, a model training phase is performed by a network device. After training an encoder and a decoder that match each other, the network device may indicate a model parameter of the encoder to a terminal device. For example, a model training phase is performed by an independent AI network element. After training an encoder and a decoder that match each other, the AI network element may send a model parameter of the encoder to a terminal device, and send a model parameter of the decoder to a network device. Then, a model inference phase corresponding to the encoder is performed in the terminal device, and a model inference phase corresponding to the decoder is performed in the network device.
Regarding claim 6 – Jin and Sajid disclose claim 1. Jin discloses wherein the predicted wireless communication data comprises one or more of: a modulation coding scheme (MCS); a downlink rank indicator; a channel quality indicator (CQI); network unit scheduler parameters; raw channel measurement of different signals; a buffer status at the UE; a decode result of any uplink or downlink grant; a scheduling request the UE has made; a load on the network unit; link adaptation parameters for an outer loop; a fairness parameter of the network unit; a traffic presence of at least another UE; a class of traffic of at least another UE; a class or priority for different UEs the network unit is currently serving; or an inference at the network unit with at least another UE, refer to paragraph [0003] - In a communication system, a network device needs to determine, based on a channel state information (channel state information, CSI) parameter, related configuration information such as a resource, a modulation and coding scheme (modulation and coding scheme, MCS), and precoding that are used to schedule a downlink data channel for a terminal device, also, refer to paragraph [0160] - For example, the CSI may include at least one of the following: a channel quality indicator (channel quality indication, CQI), a precoding matrix indicator (precoding matrix indicator, PMI), a rank indicator (rank indicator, RI), a CSI-RS resource indicator (CSI-RS resource indicator, CRI), a layer indicator (layer indicator, LI), a reference signal received power (reference signal receiving power, RSRP), a signal to interference plus noise ratio (signal to interference plus noise ratio, SINR), or the like. The signal to interference plus noise ratio may also be referred to as a signal-to-interference-plus-noise ratio.
Regarding claim 7 – Jin and Sajid disclose claim 1. Jin discloses wherein at least one of the one or more generative AI based language models is pretrained with the dataset consisting essentially of data token sequences associated with wireless protocol data, refer to paragraph [0124] - FIG. 2 is a diagram of another communication system to which a communication method according to an embodiment of this application is applicable. Compared with the communication system 100 shown in FIG. 1, the communication system 200 shown in FIG. 2 further includes an AI network element 140. The AI network element 140 is configured to perform an AI-related operation, for example, constructing a training dataset or training an AI model, also, paragraph [0125] - the network device 110 may send data related to training of the AI model to the AI network element 140, and the AI network element 140 constructs a training dataset and trains the AI model. For example, the data related to training of the AI model may include data reported by the terminal device, also, paragraph [0135] - The encoder and the decoder of the AE model are usually jointly trained and used in cooperation. It may be understood that the encoder and decoder are AI models that match each other. The encoder processes an input V to obtain a processed result z, and the decoder can decode the output z of the encoder to an expected output V′, also, paragraph [0144] - The training dataset is used to train an AI model. The training dataset may include an input of the AI model, or include the input of the AI model and a target output of the AI model. The training dataset includes one or more pieces of training data. The training data may include a training sample input to the AI model, or may include the target output of the AI model. The target output may also be referred to as a label or a label sample, also, refer to Figure 4 and paragraph [0150] - In the data collection phase, a data source (data source) is used to provide a training dataset and inference data. In the model training phase, training data (training data) provided by the data source is analyzed or trained to obtain an AI model. The AI model represents a mapping relationship between an input and an output of the model. Obtaining the AI model through learning by using a model training node is equivalent to obtaining the mapping relationship between the input and the output of the model through learning by using the training data, also, paragraph [0155] - A training dataset used for training the AI prediction model includes a training sample and a sample label. For example, the training sample is channel information measured at one or more moments, and the sample label is channel information measured at a moment after the one or more moments.
Regarding claim 8 – Jin and Sajid disclose claim 1. Jin discloses wherein at least one of the one or more generative AI based language models is pretrained using multiple datasets, each dataset consisting essentially of data token sequences associated with wireless protocol data, and each dataset corresponding to wireless protocol data belonging to a respective traffic type, refer to paragraph [0157] - In a communication system (for example, an LTE communication system or an NR communication system), a network device needs to determine, based on the CSI, configurations such as a resource, an MCS, and precoding that are used to schedule a downlink data channel for a terminal device. It may be understood that the CSI is channel information, and is information that can reflect a channel feature and channel quality. The channel information may also be referred to as a channel response. For example, the CSI may be represented by using a channel matrix. For example, the CSI includes the channel matrix, or the CSI may include a channel eigenvector, also, paragraph [0161] - For example, the terminal device may use an AI-based CSI feedback manner, to be specific, perform feedback by using an AI compression model. Specifically, the terminal device compresses and feeds back CSI by using an AI model, and the network device restores compressed CSI by using an AI model. A sequence (for example, a bit sequence) is transmitted in AI-based CSI feedback, and overheads are lower than overheads of a conventional CSI report feedback. As described above, the AI prediction model may be a two-side model. The AI prediction model and the AI compression model may be a same model. To be specific, an encoder may be configured to: predict CSI at a future moment, and compress the predicted CSI at the future moment. In other words, an output of the encoder is compressed information of the predicted CSI. The decoder may be configured to decode and restore the compressed information of the predicted CSI, to obtain the predicted CSI. Alternatively, the AI prediction model and the AI compression model may be different models, also, paragraph [0163] - CSI prediction helps alleviate a channel aging problem. The CSI prediction means that future CSI is predicted by using correlations of a channel in time domain, frequency domain, and space domain and with reference to historical CSI. The CSI prediction may be implemented in a plurality of manners, for example, linear fitting-based CSI prediction or filtering-based CSI prediction. An AI-based CSI prediction method is generated by introducing an AI technology into a wireless communication network. Compared with another manner, an AI-based CSI prediction result usually has higher prediction precision, and a small amount of historical CSI is needed. AI-based CSI prediction means that an AI prediction model is obtained through training, so that when current CSI and/or historical CSI are/is input into the AI prediction model, needed future CSI can be output, also, refer to paragraph [0174] - The second device sends indication information #1 (an example of first indication information) to the first device. The indication information #1 indicates at least one piece of channel information. When the p pieces of predicted channel information are valid, the at least one piece of channel information includes the p pieces of predicted channel information.
Regarding claim 9 – Jin and Sajid disclose claim 1. Jin discloses wherein at least one of the one or more generative AI based language models comprises an embedding layer associated with a particular network unit, wherein the one or more processors are further configured to cause the UE to: generate, by the embedding layer, an embedding vector corresponding to the particular network unit from the input token sequence, refer to paragraph [0141] - The output signal of the activation function may serve as an input of a next convolution layer. The activation function may be a sigmoid function. The neural network is a network formed by connecting many single neurons together. To be specific, an output of a neuron may be an input of another neuron. An input of each neuron may be connected to a local receptive field of a previous layer to extract a feature of the local receptive field. The local receptive field may be a region including several neurons, also, refer to paragraph [0146] - a predicted value of a current network may be compared with a target value that is actually desired, and a weight vector of each layer of the AI model is updated based on a difference between the predicted value and the target value (certainly, before a 1.sup.st update, an initialization process is usually performed, to be specific, parameters are preconfigured for all layers of the AI model) ……. adjusting a model parameter of the neural network includes adjusting at least one of the following parameters: a quantity of layers and a width of the neural network, a weight of a neuron, or a parameter in an activation function of the neuron, also, refer to paragraph [0153] - a structure parameter of the model (for example, a quantity of layers and/or a weight of the model), an input parameter of the model (for example, an input dimension and a quantity of input ports), or an output parameter of the model (for example, an output dimension and a quantity of output ports). It may be understood that the input dimension may be a size of one piece of input data. For example, when the input data is a sequence, an input dimension corresponding to the sequence may indicate a length of the sequence, paragraph [0160] - a layer indicator (layer indicator, LI).
Regarding claim 10 – Jin discloses claim 9. Jin does not explicitly disclose wherein the one or more predicted next tokens are associated with a conditional probability distribution over a vocabulary of tokens representing different types or values of wireless communication data conditioned on the embedding vector.
Sajid discloses wherein the one or more predicted next tokens are associated with a conditional probability distribution over a vocabulary of tokens representing different types or values of wireless communication data conditioned on the embedding vector, refer to paragraph [0073] - In the illustrated example, subnetworks 812, 816 are selected from the first layer 172 to process the second token 724. It is to be appreciated that a processes/mechanism used to select the subnetworks 812, 816 from the first layer 172 may be known or unknown. For instance, the subnetworks 812, 816 may be selected using a gating function or a “router network” that computes probability scores for each of the subnetworks 811-818 and selects the subnetworks 812, 816 as having the highest probability scores. In some embodiments, the subnetwork 812 includes first weights learned during training and the subnetwork 816 includes second weights learned during training that are independent of the first weights. For instance, the subnetwork 812 may include a first multi-layer perceptron and the subnetwork 816 may include a second multi-layer perceptron.
It would have been obvious to an ordinary person of skill in the art prior to the effective date of the application to combine Jin with Sajid to provide Jin with wherein the one or more predicted next tokens are associated with a conditional probability distribution over a vocabulary of tokens representing different types or values of wireless communication data conditioned on the embedding vector.
The suggestion/motivation would have been that Jin discloses – refer to paragraph [0005] - However, in an actual scenario, predicted CSI possibly cannot accurately reflect actual channel quality, affecting accuracy of a subsequent related configuration of the network device. For example, when the terminal device feeds back CSI, if an artificial intelligence (artificial intelligence, AI) prediction model is deployed on the terminal device, the terminal device may feed back, to the network device, CSI predicted by the AI prediction model.
The benefit being the AI prediction model has the potential to be more accurate – refer to paragraph [0010] - An AI model-based prediction manner is used as an example. Channel information measured at one or more historical moments is input into an AI prediction model, and the AI prediction model outputs channel information at one or more future moments.
Regarding claim 12 – Jin and Sajid disclose claim 1. Jin discloses generate by a task-specific neural network module trained using an additional dataset of wireless protocol data and corresponding performance metrics, a specific performance metric corresponding to the wireless communication data, refer to paragraph [0136] - The auto-encoder is an unsupervised learning neural network, and is characterized by using input data as label data. Therefore, the auto-encoder may also be understood as a self-supervised learning neural network. The auto-encoder may be configured to compress and restore data. For example, the encoder in the auto-encoder may compress (encode) data A to obtain data B, and the decoder in the auto-encoder may decompress (decode) the data B to restore the data A, also, refer to paragraph [0143] - For example, the type of the AI model is a neural network. The AI model in this disclosure may be a deep neural network (deep neural network, DNN). Based on a network construction mode, the DNN may include a feedforward neural network (feedforward neural network, FNN), a convolutional neural network (convolutional neural network, CNN), a recurrent neural network (recurrent neural network, RNN), and the like, also, refer to paragraph [0146] - Model training is essentially learning some features of training data from the training data. In a process of training an AI model (for example, a neural network model), because an output of the AI model is expected to be as close as possible to a predicted value that is actually desired, a predicted value of a current network may be compared with a target value that is actually desired, and a weight vector of each layer of the AI model is updated based on a difference between the predicted value and the target value (certainly, before a 1.sup.st update, an initialization process is usually performed, to be specific, parameters are preconfigured for all layers of the AI model). For example, if the predicted value of the network is large, the weight vector is adjusted to decrease the predicted value, and adjustment is continuously performed until the AI model can predict the target value that is actually desired or a value that is quite close to the target value that is actually desired, also, refer to paragraph [0156] - Then, a difference between the predicted channel information and the corresponding sample label, that is, a value of a loss function, is calculated, and a parameter of the AI prediction model is updated based on the value of the loss function, so that the difference between the predicted channel information and the corresponding sample label is minimized, that is, the loss function is minimized.
Regarding claim 14 – Jin and Sajid disclose claim 1. Jin discloses transmit, to the network unit, a capability update message indicating that the UE is configured with at least one generative AI based language model for wireless communication or link adaptation, refer to paragraph [0146] - Model training is essentially learning some features of training data from the training data. In a process of training an AI model (for example, a neural network model), because an output of the AI model is expected to be as close as possible to a predicted value that is actually desired, a predicted value of a current network may be compared with a target value that is actually desired, and a weight vector of each layer of the AI model is updated based on a difference between the predicted value and the target value (certainly, before a 1.sup.st update, an initialization process is usually performed, to be specific, parameters are preconfigured for all layers of the AI model), also, refer to Figure 4 and paragraph [0149] - For example, the actor entity may send the inference result to one or more actor objects (for example, a network device or a terminal device) for execution. For another example, the actor entity may further feed back performance of the model to the data source, to facilitate subsequent model update training, also, refer to paragraph [0156] - Then, a difference between the predicted channel information and the corresponding sample label, that is, a value of a loss function, is calculated, and a parameter of the AI prediction model is updated based on the value of the loss function, so that the difference between the predicted channel information and the corresponding sample label is minimized, that is, the loss function is minimized.
Regarding claim 15 – Jin and Sajid disclose claim 1. Jin discloses configure one or more future wireless transmissions with the network unit using the one or more communication parameters generated by the one or more generative AI-based language model, refer to paragraph [0003] - In a communication system, a network device needs to determine, based on a channel state information (channel state information, CSI) parameter, related configuration information such as a resource, a modulation and coding scheme (modulation and coding scheme, MCS), and precoding that are used to schedule a downlink data channel for a terminal device, also, refer to paragraph [0146] - Model training is essentially learning some features of training data from the training data. In a process of training an AI model (for example, a neural network model), because an output of the AI model is expected to be as close as possible to a predicted value that is actually desired, a predicted value of a current network may be compared with a target value that is actually desired, and a weight vector of each layer of the AI model is updated based on a difference between the predicted value and the target value (certainly, before a 1.sup.st update, an initialization process is usually performed, to be specific, parameters are preconfigured for all layers of the AI model) …….. For example, the AI model is a neural network, and adjusting a model parameter of the neural network includes adjusting at least one of the following parameters: a quantity of layers and a width of the neural network, a weight of a neuron, or a parameter in an activation function of the neuron, also refer to Figure 4 and paragraph [0149] - The inference data is input to the AI model, to obtain an output through the AI model, and the output is the inference result. The inference result may indicate a configuration parameter used (executed) by an actor object, and/or an operation performed by the actor object, also, refer to paragraph [0152] - For example, a model training phase is performed by a terminal device. After training an encoder and a decoder that match each other, the terminal device may send a model parameter of the decoder to a network device. For example, a model training phase is performed by a network device. After training an encoder and a decoder that match each other, the network device may indicate a model parameter of the encoder to a terminal device, also, paragraph [156] - Then, a difference between the predicted channel information and the corresponding sample label, that is, a value of a loss function, is calculated, and a parameter of the AI prediction model is updated based on the value of the loss function, so that the difference between the predicted channel information and the corresponding sample label is minimized, that is, the loss function is minimized.
Regarding claim 16 – Jin disclose A network unit for wireless communication, the network unit comprising: refer to Figure 1, also paragraph [0156] - The model training node may be a terminal device, a network device, or another network element having an AI function in a communication system.
one or more memories; and one or more processors, coupled to the one or more memories, the one or more memories configured to cause the network unit to: refer to paragraph [0093] - a computer-readable storage medium is provided. The computer-readable storage medium includes instructions. When the instructions are run by a processor,
generate an input token sequence associated with one or more of: wireless communication data between the network unit and one or more UEs that the network unit is serving, configuration data associated with the network unit or configuration data associated with the one or more UEs; refer to paragraph [0010] - the channel information may be predicted in a plurality of manners. For example, the channel information may be predicted in one or more of the following manners: linear fitting, Kalman filtering, an AI model, or the like. An AI model-based prediction manner is used as an example. Channel information measured at one or more historical moments is input into an AI prediction model, and the AI prediction model outputs channel information at one or more future moments, also, refer to paragraph [0122] - To support the AI technology in the wireless network, a dedicated AI network element or AI module may be further introduced into the network. If the AI network element is introduced, the AI network element is corresponding to an independent network element. If the AI module is introduced, the AI module may be located in a specific network element, and the corresponding network element may be a terminal device, a network device, or the like, also, refer to paragraph [0125] - the network device 110 may send data related to training of the AI model to the AI network element 140, and the AI network element 140 constructs a training dataset and trains the AI model. For example, the data related to training of the AI model may include data reported by the terminal device. The AI network element 140 may send a result of an operation related to the AI model to the network device 110, and the network device 110 forwards the result of the operation related to the AI model to the terminal device. For example, the result of the operation related to the AI model may include at least one of the following: a trained AI model, an evaluation result or a test result for the model, and the like. For example, a part of the trained AI model may be deployed on the network device 110, and the other part is deployed on the terminal device. Alternatively, the trained AI model may be deployed on the network device 110. Alternatively, the trained AI model may be deployed on the terminal device. (The examiner interprets the training dataset as the input token sequence.)
Jin does not explicitly disclose obtain one or more predicted next tokens representing predicted wireless communication data from one or more generative artificial intelligence (AI) based language models using the input token sequence, wherein the one or more generative AI based language models are pretrained with a dataset comprising data token sequences associated with wireless protocol data.
Sajid discloses obtain one or more predicted next tokens representing predicted wireless communication data from one or more generative artificial intelligence (AI) based language models using the input token sequence, wherein the one or more generative AI-based language models are pretrained with a dataset comprising data token sequences associated with wireless protocol data; refer to paragraph [0020] - A token computed based on a user input to a generative large language model may be received. A portion of the generative large language model may be identified using the token and a machine learning model trained to identify portions of the generative large language model. The portion may be written into a memory. The generative large language model may be caused to generate an output based on the user input using the portion in the memory, also, paragraph [0021] - A current token and at least one previous token generated based on a user input to a generative large language model may be received. Subnetworks within the generative large language model may be identified by processing the current token and the at least one previous token using a machine learning model. The subnetworks may be prefetched from a first memory into a second memory. The generative large language model may be caused to generate an output based on the user input using the subnetworks in the second memory, also, refer to Figure 9 and paragraph [0076] - FIG. 9 illustrates a representation of determining potential subnetworks within layers of a generative large language model 160 based on tokens, according to embodiments of the disclosure. The example shown in FIG. 9 includes a first table 902, a second table 904, and the generative large language model 160. The first table 902 is illustrated to include tokens (e.g., all of the tokens) in the vocabulary of the generative large language model 160, and claim 1.
It would have been obvious to an ordinary person of skill in the art prior to the effective date of the application to combine Jin with Sajid to provide Jin with obtain one or more predicted next tokens representing predicted wireless communication data from one or more generative artificial intelligence (AI) based language models using the input token sequence, wherein the one or more generative AI based language models are pretrained with a dataset comprising data token sequences associated with wireless protocol data.
The suggestion/motivation would have been that Jin discloses – refer to paragraph [0005] - However, in an actual scenario, predicted CSI possibly cannot accurately reflect actual channel quality, affecting accuracy of a subsequent related configuration of the network device. For example, when the terminal device feeds back CSI, if an artificial intelligence (artificial intelligence, AI) prediction model is deployed on the terminal device, the terminal device may feed back, to the network device, CSI predicted by the AI prediction model.
The benefit being the AI prediction model has the potential to be more accurate – refer to paragraph [0010] - An AI model-based prediction manner is used as an example. Channel information measured at one or more historical moments is input into an AI prediction model, and the AI prediction model outputs channel information at one or more future moments.
Jin discloses generate one or more communication parameters for future wireless communication by the network unit, based at least in part on the one or more predicted next tokens, refer to paragraph [0010] - An AI model-based prediction manner is used as an example. Channel information measured at one or more historical moments is input into an AI prediction model, and the AI prediction model outputs channel information at one or more future moments.
Regarding claim 17 – Jin and Sajid disclose claim 16. Jin discloses wherein the one or more generative AI based language models are implemented at one or more of: the network unit, or one or more distributed network units, refer to paragraph [0122] - To support the AI technology in the wireless network, a dedicated AI network element or AI module may be further introduced into the network. If the AI network element is introduced, the AI network element is corresponding to an independent network element. If the AI module is introduced, the AI module may be located in a specific network element, and the corresponding network element may be a terminal device, a network device, or the like, also, refer to Figure 2 and paragraph [0125] - In a possible implementation, the network device 110 may send data related to training of the AI model to the AI network element 140, and the AI network element 140 constructs a training dataset and trains the AI model. For example, the data related to training of the AI model may include data reported by the terminal device. The AI network element 140 may send a result of an operation related to the AI model to the network device 110, and the network device 110 forwards the result of the operation related to the AI model to the terminal device. For example, the result of the operation related to the AI model may include at least one of the following: a trained AI model, an evaluation result or a test result for the model, and the like. For example, a part of the trained AI model may be deployed on the network device 110, and the other part is deployed on the terminal device. Alternatively, the trained AI model may be deployed on the network device 110.
Regarding claim 18 – Jin and Sajid disclose claim 16. Jin discloses wherein the wireless communication data between the network unit and the one or more UEs comprises one or more of: network unit-side wireless communication data that is observed by the network unit at a current communication slot; UE-side wireless communication data that is signaled to the network unit from the one or more UEs that the network unit is serving at the current communication slot; previous wireless communication data from previous communication slots retrieved from a memory; or simulated or synthetic wireless data from system-level simulation, refer to Figure 8 and paragraph [378] - The terminal device may predict CSI at one future moment based on the m pieces of measured CSI. Specifically, the terminal device may predict the CSI at the future moment by using a prediction model. To be specific, the m pieces of measured CSI are input into the prediction model for processing, to predict the CSI at the future moment, also, refer to paragraph [0145] - In the communication field, the training dataset may include simulation data collected by using a simulation platform, or may include experimental data collected in an experiment scenario, or may include actual measurement data collected in an actual communication network,
Regarding claim 19 – Jin and Sajid disclose claim 16. Jin discloses wherein the predicted wireless communication data comprises at least one synthetic configuration data associated with at least one of the one or more UEs that the network unit is serving, refer to paragraph [0125] - In a possible implementation, the network device 110 may send data related to training of the AI model to the AI network element 140, and the AI network element 140 constructs a training dataset and trains the AI model. For example, the data related to training of the AI model may include data reported by the terminal device. The AI network element 140 may send a result of an operation related to the AI model to the network device 110, and the network device 110 forwards the result of the operation related to the AI model to the terminal device. For example, the result of the operation related to the AI model may include at least one of the following: a trained AI model, an evaluation result or a test result for the model, and the like. For example, a part of the trained AI model may be deployed on the network device 110, and the other part is deployed on the terminal device. Alternatively, the trained AI model may be deployed on the network device 110, also, refer to paragraph [0136] - The auto-encoder is an unsupervised learning neural network, and is characterized by using input data as label data. Therefore, the auto-encoder may also be understood as a self-supervised learning neural network. The auto-encoder may be configured to compress and restore data. For example, the encoder in the auto-encoder may compress (encode) data A to obtain data B, and the decoder in the auto-encoder may decompress (decode) the data B to restore the data A, also, refer to paragraph [0144] - The training dataset is used to train an AI model. The training dataset may include an input of the AI model, or include the input of the AI model and a target output of the AI model. The training dataset includes one or more pieces of training data. The training data may include a training sample input to the AI model, or may include the target output of the AI model. The target output may also be referred to as a label or a label sample, also, refer to paragraph [0145] - In the communication field, the training dataset may include simulation data collected by using a simulation platform, or may include experimental data collected in an experiment scenario, or may include actual measurement data collected in an actual communication network, also, refer to paragraph [0152] -Training processes for different models may be deployed on different devices or nodes, or may be deployed on a same device or node. Inference processes for different models may be deployed on different devices or nodes, or may be deployed on a same device or node. For example, a model training phase is performed by a terminal device. After training an encoder and a decoder that match each other, the terminal device may send a model parameter of the decoder to a network device. For example, a model training phase is performed by a network device. After training an encoder and a decoder that match each other, the network device may indicate a model parameter of the encoder to a terminal device. For example, a model training phase is performed by an independent AI network element. After training an encoder and a decoder that match each other, the AI network element may send a model parameter of the encoder to a terminal device, and send a model parameter of the decoder to a network device. Then, a model inference phase corresponding to the encoder is performed in the terminal device, and a model inference phase corresponding to the decoder is performed in the network device.
Regarding claim 20 – Jin and Sajid disclose claim 16. Jin discloses wherein the predicted wireless communication data comprises one or more of: a modulation coding scheme (MCS); a downlink rank indicator; a channel quality indicator (CQI); network unit scheduler parameters; raw channel measurement of different signals; a buffer status at the UE; a decode result of any uplink or downlink grant; a scheduling request the UE has made; a load on the network unit; link adaptation parameters for an outer loop; a fairness parameter of the network unit; a traffic presence of at least another UE; a class of traffic of at least another UE; a class or priority for one or more UEs the network unit is currently serving; or an inference at the network unit with at least another UE, refer to paragraph [0003] - In a communication system, a network device needs to determine, based on a channel state information (channel state information, CSI) parameter, related configuration information such as a resource, a modulation and coding scheme (modulation and coding scheme, MCS), and precoding that are used to schedule a downlink data channel for a terminal device, also, refer to paragraph [0160] - For example, the CSI may include at least one of the following: a channel quality indicator (channel quality indication, CQI), a precoding matrix indicator (precoding matrix indicator, PMI), a rank indicator (rank indicator, RI), a CSI-RS resource indicator (CSI-RS resource indicator, CRI), a layer indicator (layer indicator, LI), a reference signal received power (reference signal receiving power, RSRP), a signal to interference plus noise ratio (signal to interference plus noise ratio, SINR), or the like. The signal to interference plus noise ratio may also be referred to as a signal-to-interference-plus-noise ratio.
Regarding claim 21 – Jin and Sajid disclose claim 16. Jin discloses wherein at least one of the one or more generative AI based language models is pretrained with the dataset consisting essentially of data token sequences associated with wireless protocol data, refer to paragraph [0124] - FIG. 2 is a diagram of another communication system to which a communication method according to an embodiment of this application is applicable. Compared with the communication system 100 shown in FIG. 1, the communication system 200 shown in FIG. 2 further includes an AI network element 140. The AI network element 140 is configured to perform an AI-related operation, for example, constructing a training dataset or training an AI model, also, paragraph [0125] - the network device 110 may send data related to training of the AI model to the AI network element 140, and the AI network element 140 constructs a training dataset and trains the AI model. For example, the data related to training of the AI model may include data reported by the terminal device, also, paragraph [0135] - The encoder and the decoder of the AE model are usually jointly trained and used in cooperation. It may be understood that the encoder and decoder are AI models that match each other. The encoder processes an input V to obtain a processed result z, and the decoder can decode the output z of the encoder to an expected output V′, also, paragraph [0144] - The training dataset is used to train an AI model. The training dataset may include an input of the AI model, or include the input of the AI model and a target output of the AI model. The training dataset includes one or more pieces of training data. The training data may include a training sample input to the AI model, or may include the target output of the AI model. The target output may also be referred to as a label or a label sample, also, refer to Figure 4 and paragraph [0150] - In the data collection phase, a data source (data source) is used to provide a training dataset and inference data. In the model training phase, training data (training data) provided by the data source is analyzed or trained to obtain an AI model. The AI model represents a mapping relationship between an input and an output of the model. Obtaining the AI model through learning by using a model training node is equivalent to obtaining the mapping relationship between the input and the output of the model through learning by using the training data, also, paragraph [0155] - A training dataset used for training the AI prediction model includes a training sample and a sample label. For example, the training sample is channel information measured at one or more moments, and the sample label is channel information measured at a moment after the one or more moments.
Regarding claim 22 – Jin and Sajid disclose claim 16. Jin discloses wherein at least one of the one or more generative Al based language models is pre-trained using multiple datascts, each dataset consisting essentially of data token sequences associated with wireless protocol data, and each dataset corresponding to wireless protocol data belonging to a respective traffic type, refer to paragraph [0157] - In a communication system (for example, an LTE communication system or an NR communication system), a network device needs to determine, based on the CSI, configurations such as a resource, an MCS, and precoding that are used to schedule a downlink data channel for a terminal device. It may be understood that the CSI is channel information, and is information that can reflect a channel feature and channel quality. The channel information may also be referred to as a channel response. For example, the CSI may be represented by using a channel matrix. For example, the CSI includes the channel matrix, or the CSI may include a channel eigenvector, also, paragraph [0161] - For example, the terminal device may use an AI-based CSI feedback manner, to be specific, perform feedback by using an AI compression model. Specifically, the terminal device compresses and feeds back CSI by using an AI model, and the network device restores compressed CSI by using an AI model. A sequence (for example, a bit sequence) is transmitted in AI-based CSI feedback, and overheads are lower than overheads of a conventional CSI report feedback. As described above, the AI prediction model may be a two-side model. The AI prediction model and the AI compression model may be a same model. To be specific, an encoder may be configured to: predict CSI at a future moment, and compress the predicted CSI at the future moment. In other words, an output of the encoder is compressed information of the predicted CSI. The decoder may be configured to decode and restore the compressed information of the predicted CSI, to obtain the predicted CSI. Alternatively, the AI prediction model and the AI compression model may be different models, also, paragraph [0163] - CSI prediction helps alleviate a channel aging problem. The CSI prediction means that future CSI is predicted by using correlations of a channel in time domain, frequency domain, and space domain and with reference to historical CSI. The CSI prediction may be implemented in a plurality of manners, for example, linear fitting-based CSI prediction or filtering-based CSI prediction. An AI-based CSI prediction method is generated by introducing an AI technology into a wireless communication network. Compared with another manner, an AI-based CSI prediction result usually has higher prediction precision, and a small amount of historical CSI is needed. AI-based CSI prediction means that an AI prediction model is obtained through training, so that when current CSI and/or historical CSI are/is input into the AI prediction model, needed future CSI can be output, also, refer to paragraph [0174] - The second device sends indication information #1 (an example of first indication information) to the first device. The indication information #1 indicates at least one piece of channel information. When the p pieces of predicted channel information are valid, the at least one piece of channel information includes the p pieces of predicted channel information.
Regarding claim 23 – Jin and Sajid disclose claim 16. Jin discloses wherein at least one of the one or more generative AI based language models comprises an embedding layer that is specific to the network unit, wherein the one or more processors are further configured to cause the network unit to: generate, by the embedding layer, an embedding vector that is specific to the network unit, from the input token sequence, refer to paragraph [0141] - The output signal of the activation function may serve as an input of a next convolution layer. The activation function may be a sigmoid function. The neural network is a network formed by connecting many single neurons together. To be specific, an output of a neuron may be an input of another neuron. An input of each neuron may be connected to a local receptive field of a previous layer to extract a feature of the local receptive field. The local receptive field may be a region including several neurons, also, refer to paragraph [0146] - a predicted value of a current network may be compared with a target value that is actually desired, and a weight vector of each layer of the AI model is updated based on a difference between the predicted value and the target value (certainly, before a 1.sup.st update, an initialization process is usually performed, to be specific, parameters are preconfigured for all layers of the AI model) ……. adjusting a model parameter of the neural network includes adjusting at least one of the following parameters: a quantity of layers and a width of the neural network, a weight of a neuron, or a parameter in an activation function of the neuron, also, refer to paragraph [0153] - a structure parameter of the model (for example, a quantity of layers and/or a weight of the model), an input parameter of the model (for example, an input dimension and a quantity of input ports), or an output parameter of the model (for example, an output dimension and a quantity of output ports). It may be understood that the input dimension may be a size of one piece of input data. For example, when the input data is a sequence, an input dimension corresponding to the sequence may indicate a length of the sequence, paragraph [0160] - a layer indicator (layer indicator, LI).
Regarding claim 24 – Jin and Sajid disclose claim 16. Jin does not explicitly disclose wherein the one or more predicted next tokens are associated with a conditional probability distribution over a vocabulary of tokens representing different types or values of wireless communication data conditioned on the embedding vector.
Sajid discloses wherein the one or more predicted next tokens are associated with a conditional probability distribution over a vocabulary of tokens representing different types or values of wireless communication data conditioned on the embedding vector, refer to paragraph [0073] - In the illustrated example, subnetworks 812, 816 are selected from the first layer 172 to process the second token 724. It is to be appreciated that a processes/mechanism used to select the subnetworks 812, 816 from the first layer 172 may be known or unknown. For instance, the subnetworks 812, 816 may be selected using a gating function or a “router network” that computes probability scores for each of the subnetworks 811-818 and selects the subnetworks 812, 816 as having the highest probability scores. In some embodiments, the subnetwork 812 includes first weights learned during training and the subnetwork 816 includes second weights learned during training that are independent of the first weights. For instance, the subnetwork 812 may include a first multi-layer perceptron and the subnetwork 816 may include a second multi-layer perceptron.
It would have been obvious to an ordinary person of skill in the art prior to the effective date of the application to combine Jin with Sajid to provide Jin with wherein the one or more predicted next tokens are associated with a conditional probability distribution over a vocabulary of tokens representing different types or values of wireless communication data conditioned on the embedding vector.
The suggestion/motivation would have been that Jin discloses – refer to paragraph [0005] - However, in an actual scenario, predicted CSI possibly cannot accurately reflect actual channel quality, affecting accuracy of a subsequent related configuration of the network device. For example, when the terminal device feeds back CSI, if an artificial intelligence (artificial intelligence, AI) prediction model is deployed on the terminal device, the terminal device may feed back, to the network device, CSI predicted by the AI prediction model.
The benefit being the AI prediction model has the potential to be more accurate – refer to paragraph [0010] - An AI model-based prediction manner is used as an example. Channel information measured at one or more historical moments is input into an AI prediction model, and the AI prediction model outputs channel information at one or more future moments.
Regarding claim 26 – Jin and Sajid disclose claim 16. Jin discloses wherein the one or more processors are further configured to cause the network unit toto: generate, by a task-specific neural network module trained using an additional dataset of wireless protocol data and corresponding performance metrics, a specific performance metric corresponding to the wireless communication data between the network unit and the one or more UEs, refer to paragraph [0136] - The auto-encoder is an unsupervised learning neural network, and is characterized by using input data as label data. Therefore, the auto-encoder may also be understood as a self-supervised learning neural network. The auto-encoder may be configured to compress and restore data. For example, the encoder in the auto-encoder may compress (encode) data A to obtain data B, and the decoder in the auto-encoder may decompress (decode) the data B to restore the data A, also, refer to paragraph [0143] - For example, the type of the AI model is a neural network. The AI model in this disclosure may be a deep neural network (deep neural network, DNN). Based on a network construction mode, the DNN may include a feedforward neural network (feedforward neural network, FNN), a convolutional neural network (convolutional neural network, CNN), a recurrent neural network (recurrent neural network, RNN), and the like, also, refer to paragraph [0146] - Model training is essentially learning some features of training data from the training data. In a process of training an AI model (for example, a neural network model), because an output of the AI model is expected to be as close as possible to a predicted value that is actually desired, a predicted value of a current network may be compared with a target value that is actually desired, and a weight vector of each layer of the AI model is updated based on a difference between the predicted value and the target value (certainly, before a 1.sup.st update, an initialization process is usually performed, to be specific, parameters are preconfigured for all layers of the AI model). For example, if the predicted value of the network is large, the weight vector is adjusted to decrease the predicted value, and adjustment is continuously performed until the AI model can predict the target value that is actually desired or a value that is quite close to the target value that is actually desired, also, refer to paragraph [0156] - Then, a difference between the predicted channel information and the corresponding sample label, that is, a value of a loss function, is calculated, and a parameter of the AI prediction model is updated based on the value of the loss function, so that the difference between the predicted channel information and the corresponding sample label is minimized, that is, the loss function is minimized.
Regarding claim 31 – Please use claim 1 for the rejection of claim 31.
Regarding claim 32 - Please use claim 16 for the rejection of claim 32.
Allowable Subject Matter
Claims 11, 13, 25, and 27-30 are allowable over the prior art of record.
Reasons for Allowance
The following is an examiner’s statement of reasons for allowance: Applicants have claimed uniquely distinct features in the application, which are not found in the prior art, either singularly or in combination. The independent claims identify the following uniquely distinct features:
I. The primary reason for the allowance of the claims are the inclusion of the limitation in the claims which are not found in the prior art references. The following claim elements “auto-regressively generate, by the one or more generative Al based language models, one or more updated predicted next tokens conditioned on the next input sequence of tokens
” together with the other elements are the reasons for allowance.
1. Regarding claims 11 and 25 – sample the one or more predicted next tokens from the vocabulary of tokens according to the conditional probability distribution; combine the sampled one or more predicted next tokens into a next input sequence of tokens; and auto-regressively generate, by the one or more generative Al based language models, one or more updated predicted next tokens conditioned on the next input sequence of tokens.
2. Regarding claim 13 and 27 – wherein the task-specific neural network module is trained in conjunction with at least one of the one or more generative AI based language models to produce a training output while the at least one of the one or more generative Al based language models remain frozen during backpropagation.
The closest prior art, either singularly or in combination, fail to anticipate or render the above limitations obvious.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
1. Galvin (US 2025/0363362 A1) discloses dynamically-encoded agent network for optimized deep learning.
2. Peng et al. (US 2026/0093932 A1) discloses systems and methods for providing self-improving AI models.
3. Fakoorian et al. (US 2024/0406036 A1) discloses channel estimation using AI.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to John Pezzlo whose telephone number is (571) 272-3090. The examiner can normally be reached on Monday to Friday from 8:30 AM to 5:00 PM.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Ayman A. Abaza, can be reached at telephone number (571) 270-0422. The fax phone number for the organization where this application or proceeding is assigned is (571) 273-8300.
Information regarding the status of an application may be obtained from Patent Center and the Private Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from Patent Center or Private PAIR. Status information for unpublished applications is available through Patent Center and Private PAIR to authorized users only. Should you have questions about access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free).
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) Form at https://www.uspto.gov/patents/uspto-automated- interview-request-air-form .
John Pezzlo
25 September 2026
/John Pezzlo/
Primary Examiner, Art Unit 2465B
